Yet another disasm.py speedup

[AngheloAlf]

Change disasm.py to use rabbitizer as the instruction decoder instead of the bundled mips_isa.py file.

The advantage of using this package instead of what we are doing now is this package is a fast MIPS instruction decoder fully written in C, so runtime and memory usage are greatly improved.
For example, running the disasm.py script on current master:

$ time python3 tools/disasm/disasm.py -j 1 --reg-names=o32 --all

...

real    1m1.855s
user    1m1.200s
sys    0m4.860s

This PR:

$ time python3 tools/disasm/disasm.py -j 1 --reg-names=o32 --all

...

real    0m32.606s
user    0m30.492s
sys    0m4.293s

(I couldn't get proper memory usage measurements because the multithreading stuff messed up with them. lmk if anybody knows a proper way to measure it)

I cleaned up and de-duplicated some code. It is a bit more readable imo.

I also took this opportunity to clean up the requirements.txt file and remove what seems to be dead dependencies and add diff.py dependencies and rabbitizer

[hensldm]

I'm also checking it out to do some testing

[hensldm]

Seems to work fine for me. Always enjoy the speed ups :) .

[EllipticEllipsis]

Works on Mac, code looks good from what I can see, but there's a few issues with the surroundings.

[EllipticEllipsis]

I wish we could just subrepo it instead, this basically counters all new timesave.

[AngheloAlf]

The package is written in C, so people would require to build it themselves. They would require to install the python dev package to build it too. People on Windows (cygwin, not wsl) would require the VisualStudio toolchain just to build it. Too much trouble imo.

Also, this step only adds 2 extra seconds on the jenkins side

[hensldm]

Yeah subrepo doesn't really help since Jenkins would have to build it anyways, which is almost certainly going to be longer than just installing a few python packages.

[EllipticEllipsis]

This is a subrepo, right? So this should be PR'd to that repo. And the permuter might need it too. And maybe decomp.me?

[AngheloAlf]

No. As every time this topic comes up, the asm processor is not a subrepo... for reasons I can never remember.

If using this register alias is an issue I can disable it on rabbitizer's configuration

[hensldm]

#278 (review)

[EllipticEllipsis]

I think we can probably live without this, I expect mips2c m2c wouldn't like it either.

[AngheloAlf]

Removed the use of $FpcCsr for now instead