I think that for instrumentation versus computationally heavy theory, the number of positions both in academia and in industry are comparable. In many ways computational theory is an area in which the computer is another instrument. In astronomy, instrumentation people work on hardware and computational theorists work on software, and the numbers are roughly comparable.
That's a different question. One thing that isn't appreciated is that universities and research institutes hire rather large numbers of staff are non-tenured. If you go into a telescope, you'll find that most people that work on telescopes are non-tenured research scientists or technicians. The same holds true for computer centers.
One thing about computational theory and instrumentation is that the job prospects are considerably better than "pencil and paper" theory because of the large number of non-tenured staff positions.
Yup. If you worked at a big supercomputing center, (NCSA or UCSD) then things would be different. Also, Europe has had a lot different priorities than the US.
One reason astrophysics computation is important in the US (as well in China and Russia) has to do with making sure that the bombs still work. People have come to the conclusion that it would be better to ban nuclear testing worldwide, because if you ban nuclear testing, then it makes it harder for Iran, Pakistan, and North Korea to get the practical experience in building nuclear weapons. Once you can no longer blow up bombs, then you have to test the weapons via simulation, which means that the people that run things make sure that a lot of money goes into supercomputing, because it would be a really bummer if there is a major international crisis, and North Korea or Iran is willing to gamble that American H-bombs just don't work.
So in the US, there are a lot of jobs in defense. This matters even if you can't or don't want to build H-bombs, because anyone that gets hired at Los Alamos, is one less person looking for a non-defense related job. Also, I have a suspicion that people that know how to make a telescope that looks up, also get jobs building telescopes that look down.
One other suggestion is to take the *opposite* of what you are interested in. If you are interested in computation theory, it would make more sense if you took a course on instrumentation, and if you are interested in instrumentation, it makes more sense if you took a course on theory. To be a competent theorist, you need to know basic instrumentation (otherwise you end up making theories that are untestable and hence useless) and to be a competent experimentalist, you need to know basic theory (otherwise, you have no idea what you are looking for).
The reason for this is that if you go instrumentation, you are going to be spending years and years doing it, whereas if you go theory, you are going to be spending years and years doing that. Your exposure to the "other side" is going to be limited, and it may be a good idea to get that now.
Also, it will help with grad school admissions to be less specialized than more specialized.