Not being in the industry I am not willing to go hard facts but I reckon I can take your questions.
Also programmers: those coming from physics, maths or electronics for stuff like this are probably better than those come from straight up computing (on the flip side I would probably not hire an electrical engineer to manage my corporate IT setup). Games courses.... I have not researched them in depth but I do know they are anything but a sure fire way to end up in the industry and what I hear says they do not confer much, if any, advantage so I would not dwell on it.
Now for physics: All physics (and science in general) is ultimately a best fit (be it simplest or most accurate) toolkit for a given scenario- this leads to the the unified theory of physics aka the holy grail or science.
Now computers are in the grand scheme of things very slow so if you can get away with saying acceleration due to gravity= 10ms^-2 rather than the standard field strength equation* or worse the "dark matter vs gravity (MOND)" stuff you will.
*even if you do implement it you are probably going to simplify it again but using big mass/coefficients and the like to net the "same numbers" without the need for massive distances and masses and just to tie it into computing the need for big/small floating point numbers.
This applies to light (high end light physics is horrific to calculate for a single beam with but a few interactions let alone many with many*), water (our supercomputers are as yet unable to do turbulence all that well if at all even if we have the science), friction (the coefficient changes with material, speed, forces, wear, surface and much more) and just about every other area of physics. If you are going to play the programmer your job will be to pick and choose the physics necessary for the job and encode in a way that the computer can crunch numbers for; this is where the difference between schoolboy physics and game physics lies and why I would argue you need to be on top of your game in both physics and computing (nicely both make extensive use of matrices). This line of thought also goes for AI (ever seen a simple pattern in the AI of a game?), rendering techniques (some of the more recent stuff will only render what you "see" rather than everything to a cutoff point including stuff that is blocked) and as a nice hybrid of computing and physics as well as and example of it all see some of the work that went into new methods for rendering fog.
*lighting is big right now in 3d/games:
http://www.3dgraphics101.com/3dgraphics102/
Case in point how many games fail to use Newton's first law (in motion.... stays in motion) if the motion you have is via a third party (two examples from my head: duke nukem 3d; train level, press jump and Tony Hawk's 4; jump in the back of truck and ollie as it stops/moves) yet aside from me trying to break/test the engine work fine.
By all means do animation but I suggest you do it as a night school/distance learning sort of thing instead.
Disclosure: I am an engineer so physics is naturally somewhat important to me but having pulled apart more than my fair share of games it seems pretty important there too.
