Here's a big problem with conventional nuclear bombs: they don't scale down. There's a minimum size dictated by the fissile material you're using. A pure plutonium-239 device would need 11kg of plutonium, and yield an equivalent of 10-20 tonnes of TNT (have a read up on the W54 warhead for a real-world device of that yield). That's about 40GJ, which is quite a lot of energy to expend on a single machining operation (and it is also quite an inefficient way to use up all that hard-to-make plutonium, which could be better used as the first stage of a much more powerful teller-ulam style thermonuclear device). Further inefficiencies will waste a fairly big chunk of that energy, but it is still a hell of a bang.
Next, remember that a casaba howitzer focuses a great deal of its energy in the desired direction, but not all of it. This means quite a lot ends up going in other directions, meaning you can't deploy your nuclear hammer too close to any other infrastructure that isn't very well protected. You probably want to avoid doing it in low orbits, too, because EMP is likely to annoy the neighbours and might annoy your own people, too.
Now, the stuff actually hitting the target will be a mixture of x-rays, neutrons, electrons and light nuclei. Each of those will affect the target in slightly different ways, and they won't all arrive at quite the same time. This is something of an inconvenience, engineering wise.
Finally, the nuclear blast hitting the target isn't going to neatly melt a hole through it. It'll heat up a comparatively shallow (have a look at ToughSF's calculations for some idea of depth; they're not correct, but they're in the right ballpark) section of the surface very rapidly, which may cause it to vapourise explosively, melt or merely suffer all sorts of interesting thermal and radiation effects. It won't have a neatly defined edge-of-beam either; power levels will drop off sharply outside of the bullseye, but given the scales involved ("ground zero" will probably be metres across at the target) there will be an obvious transition at the target which will make for ugly burning, scarring or fracturing none of which really imply "precision". See Itmauve's answer above for what a messy machining process looks like at small scales; now imagine it about 10m across, red hot and possibly radioactive.
So, it'd be expensive, inefficient, clumsy, messy and probably massive overkill. Fine for weapons, not fine for anything else.
You might try to fix things by making some gigantic electromagnetic focussing systems to collimate the nuclear blast still further, but at that point you may as well just build a simpler particle beam system that gives you much better control of beam parameters with none of the messy side effects of having an actual nuclear bomb going off.
Stick to lasers and more conventional particle beam systems for precision work.