Add back the old implementation of vfpu_sin/cos/sincos.

[hrydgard]

I hate to do this, but here we go.

Reverts #13526 (#13705 (Cho Aniki Zero) and #13671 (Hajime no Ippo) broke), but keeps the new implementation too, specifically for Hitman Reborn Battle Arena 2, see #13605 (uses a compat.ini flag).

Temporary measure to avoid regressions in 1.11.

Hope to more accurately reverse engineer these operations in the future.

[somepunkid]

Forgot to add this, but this temp fix also applies to ULJS00157 - Hitman Reborn Battle Arena (the first game). Same exact issues with both games, perhaps the same variable can be added to the compat for now for future builds.

[hrydgard]

@somepunkid thanks, applied.

[unknownbrackets]

I had done some initial testing at some point and it seemed like the least significant 8 bits of the results were always zero, but maybe I'm mixing my memory with the square root.

Probably the best option is to model this with a lot of results and figure out what gives us the most matches.

-[Unknown]

[somepunkid]

I had done some initial testing at some point and it seemed like the least significant 8 bits of the results were always zero, but maybe I'm mixing my memory with the square root.

Probably the best option is to model this with a lot of results and figure out what gives us the most matches.

-[Unknown]

If this is attempted, or build testing is required, I can test variables in-game. Thanks to you both, as always :>