Improve PSP VFPU instruction parsing & add tests

[tclamb]

I ran into a few issues with VFPU functions while iterating on mods. It seems like tokenizer changes may have broken a few of these instructions at some point? This commit fixes some discrepancies I found with the pspdev vfpu documentation, aligns prefix syntax with binutils/MWCC, and adds tests to prevent future regressions.

• reworked vfim.s to accept float & int literals (previously took integer representations of IEEE754 float32s, e.g. 0x3f800000 for 1.0)
• removed support for the invalid vsbn.{p,t,q} instructions
• fixed vt size for vscl.{p,t,q} (is always scalar, regardless of vs & vd)
• fixed vd size for the various pack/unpack instructions
• renamed vuc2i.s to vuc2ifs.s
• fixed various issues with vpfx{s,t,d} parsing
  • the sequence parser is greedy, so 1 would match before fractions like 1/3 (fixed)
  • numeric literals were not matched at all since the parser only expected identifier tokens (fixed)
  • requiring the operands to be wrapped with [/] breaks compatibility with other tooling (removed)
  • m is a write mask operand that prevents writing to a certain destination slot, not a modifier on a saturation operation (fixed)
  • 0-1/-1-1 syntax for the vpfxd saturation operations [0:1]/[-1:1] is not supported elsewhere, and the unbracketed 0:1/-1:1 syntax is not supported by MWCC (removed)
• fixed 0 support in vrot.{p,t,q} (had expected an identifier token, not an integer)

I used MWCC to generate expected output for the added/modified tests.

[Kingcom]

Thanks a lot! This looks very nice. The original implementation of those instructions definitely didn't get enough testing... I think I mostly tried to copy what I saw in PPSSPP at the time.

Just one small thing, can you add a quick note at the top of the changelog in the Readme summarizing the breaking changes, maybe with a link to the documentation? Under a Current Development Version header or something like that.