Implement IMUL, PCNT and CONT shader instructions, fix FFMA32I and HFMA32I

[gdkchan]

A bunch of shader fixes here:

• Implement the IMUL (integer multiply) shader instruction. Used by MelonDS Switch port and possible other homebrew. Not sure if commercial games uses this shader instruction.
• Implement PCNT and CONT shader instructions. Those are similar to existing PBK/BRK and SSY/SYNC pairs. Also used by MelonDS Switch port and possibly other homebrew, AFAIK the NVIDIA compiler never uses this instruction.
• Fix FMUL32I instruction implementation. Third operand should use the destination register, not "SrcC" as it does not exist for this instruction. (Regression from Rewrite shader decoding stage #2698).
• Fix a issue similar to the above one for HFMA32I aswell, but this one was also missing from the instruciton table so I have added it too. (Technically not a regression as this instruction was not implemented before).

Testing on games is welcome. MelonDS is tricky to test right now as it requires other changes to work, plus only the newer version uses GPU rendering, the older ones are software rendered and won't be affected by those changes.

[github-actions]

Download the artifacts for this pull request:

• ryujinx-headless-sdl2-Release-1.0.0+4903ce6-linux_x64
• ryujinx-headless-sdl2-Release-1.0.0+4903ce6-osx_x64
• ryujinx-headless-sdl2-Release-1.0.0+4903ce6-win_x64
• ryujinx-Release-1.0.0+4903ce6-linux_x64
• ryujinx-Release-1.0.0+4903ce6-osx_x64
• ryujinx-Release-1.0.0+4903ce6-win_x64

Only for Developers

• ryujinx-Debug-1.0.0+4903ce6-linux_x64
• ryujinx-Debug-1.0.0+4903ce6-osx_x64
• ryujinx-Debug-1.0.0+4903ce6-win_x64
• ryujinx-headless-sdl2-Debug-1.0.0+4903ce6-linux_x64
• ryujinx-headless-sdl2-Debug-1.0.0+4903ce6-osx_x64
• ryujinx-headless-sdl2-Debug-1.0.0+4903ce6-win_x64

[riperiperi]

Looking good, nice work.