Look at Real World Examples of What Works
If it looks like a duck and quacks like a duck then its a duck, or in your case, if it looks like infantry and goes bang like infantry, then it is infantry. But this would make for a lame answer; so, instead of exactly answering how it can be used (since you are already clearly looking to fill an infantry role) I will instead answer how you can refine this design to fill its intended role much better with the tech you already have.
How could simple but cheap robots be used in combat?
First, if your goal is a cheap, then you should consider redesigning a few aspects of them. While the real world already has the technology to do more or less what you are asking for, actual militaries are not moving in the direction of humanoid bots with many hardpoints because such bots simply do not add any utility that other, cheaper body plans could achieve.
If you want multiple weapon systems, I would suggest a single integrated hard point with a weapon system more like an m-16/m-203 combo. Or if you think in terms of tanks, like how the M3 Bradley has a cannon and missile system on the same turret. It reduces cost, weight, and complexity without really giving up any practical combat potential. Also, if you design the hardpoint more like a tank turret, it means much better armor, range of motion, and recoil handling, than a humanoid arm/shoulder like hardpoint. Since these bots are designed to operate in squads though, I would suggest an interchangeable turret designed to pack operation specific weapon systems. While there may be cases where a multi-role turret would be ideal, sometimes it is better to pack one weapon that fills its desired role very well than several weapons that are only okay at their jobs. So you could, if need be, take out the assault riffle/grenade launcher combo on some bots and instead give them one much heavier weapon like a heavy machine gun, an anti-tank missile launcher, or flame thrower. If you focus on many smaller hardpoints, no one of them will have the same maximum capacity for this.
Secondly, wheels are better than feet in most cases. They are more energy efficient, more resistant to damage, and much faster. If you need your robots to go indoors sometimes, give them a convertible wheel system like the Hyundai Elevate, but these will probably only need be on some of your droids. Walkable wheels are going to be more expensive and less reliable So, if your droids come in squads of 25, perhaps some are little more than mini-person sized tanks, while others are more specialized infiltration units that can convert to climb stairs when the need arises.
I would also only give arms to your infiltration variety but only 1. Nothing a robot needs to do when sieging a building actually requires more than one hand, so don't waste the parts on two. In rare cases where 2 hands are needed for something, they can work together. You can see this price saving feature in the design of Bostin Dynamic's Spot which has been proving the ability of 1-armed robots to get around urban environments like this just fine for the past couple of years.
For armor, you are describing STANAG 4569 Level 4 armor which is pretty heavy stuff for what is meant to be a cheap droid. Such a droid would need ~31 mm of ceramic/composite armor to block a BMG-50 round. At ~90kg/m^2 your droids would need to invest not just in armor weight, but in a hefty engine
too. If we consider a robot with just 3-4m^2 of armor and that tanks are normally 35-50% armor by weight, this means we are looking at robot that weighs as much as a small sedan. This could pose a significant problem moving such a small bot through buildings or soft terrain. I would instead suggest 16-25mm ceramic/composite armor over most of the body for level 2-3 protection. Level-2 will still stop most standard infantry rounds, and level-3 will stop most Armor Piercing infantry rounds, all at a much lower weight and cost. I would only put anything as heavy as level-4 on the front ends of the non-infiltration units since that is by far the most likely place for you to have to take anything as heavy as a 50-BMG like round. Of course, if you use sloped armor, you may still be able to get away with a thinner, lighter armor and still be able to stop a BMG-50 round, but only assuming it hits you at a "good" angle. With selective armoring as described above, you could probably reduce the average weight per unit down to something in the 300-500kg range which means cheaper to build, cheaper to transport, cheaper to operate, better for entering buildings, and still able to handle most of the threats it realistically would need to face.
As for control systems: consider something like the IDF Jaguar. This Hybrid AI weapons system is basically just what you are asking for. A single operator can control multiple Jaguars by setting way points and letting the AI do all the stuff that normally takes a full person's focus. Since their AI is not quite good enough to ensure an ethical kill, the target confirmation is always done by a remote operator. Otherwise, they can pretty much fill the role of normal infantry with only one operator able to control several of them.
Sizing them to make them fit?
Your droids have to fit things in 3ways.
1: It needs to fit through most standard international doors which are 2.5ft (76cm) wide.
2: It needs to be able to be small and light enough to fit 25 to a standard freight container so you that you don't need to make special accommodations to transport a squad of them. Your standard freight container has an internal cross-section of 7'8.125" x 7'9.525" (234x238cm), and come in length increments of 10, 20, & 40ft (305, 610, & 914 cm), and your average freight truck can pull 40 tons.
3: It needs to have the verticality to do what people do.
So, if you keep them just under 23" (58cm) wide, then you can pack them into a standard container 4 to a row and they can fit through standard international doors with a few inches to spare. Length will be more needed to make sure you can fit the needed power/propulsion/electronics systems this thing will need; so, longer is better, but too long and you will again have issues with moving around indoors; so, I would cap this at 4' (117cm). For this gives you enough room to fit up to 40 in a 40' truck... or in your case, 25 plus some room for your human operations control station. Going back to my armor/weight estimates, this also puts you at a weight of <12.5 metric tons of bots meaning you have >25 tons of weight capacity left to meet the needs of your control station and logistics which should be plenty.
As for being 6'3" tall as defined by the OP, it would have a sensor boom similar to the IDF Jaguar; so, it can see over things a person can see over but it's actual turret would be at about hip height when standing in it's walk mode, or in it's drive mode it would have an even lower profile mimicking a person in a more prone position. This way it has the advantages of both a vertical human body plan and horizontal vehicle.
What do you get when you add up all these features?
What you have is a weapons platform that, if mass produced, would have a pretty small price tag as far as military expenditures go. It can do all the things a foot soldier can do: travel through any terrain, infiltrate building, etc, but has much better armor, advanced sensors, better aiming, better carry capacity, can move much faster over even terrain, and does not endanger the life of a human combatant. Because of the semiautonomous nature of the AI, anything AIs do better than humans can be done by the AI, anything humans do better can be done by the humans, all while you run a learning AI to watch the human operators behaviors until the AI gets better at those things too. In short, if you want to make a near-future semi-autonomous infantry replacement bot, all the pieces you need to make it happen already exist, you just need to put them together.