Surface module compiling in SDL3

[Starbuck5]

This PR consists of 4 commits, and it gets the surface module to compile with SDL3. It still fails lots of tests when run in SDL3 mode, but there is lots of SDL-function porting to be done with int->bool stuff.

Commit 1 focuses on Surface.c

This commit does not change the SDL2 codepath at all, except for a cliprect thing similar to what I've done in other PRs.

The RLEaccel stuff is confusing, even digging through the SDL2 source it's hard to understand. Our code uses surf->flags & SDL_RLEACCEL to check, interchangeably with SDL_SurfaceHasRLE(surf), but if you check the source SDL_SurfaceHasRLE does not check the same flags. However, in testing it seems to be equivalent, somehow.

I also just declined to port a whole section that covers a blitting edge case, we'll get to it later, for now I think it will be valuable to have people able to compile at all.

EDIT: also added a commit 5 for surface.c.

Commit 2 and 3 focus on filling and blitting respectively

There are a lot of lines changed but it's just passing formats around, nothing complex

Commit 4 makes intrinsics compile properly

I took Ankith's prior work on this and moved it up to _surface.h, as well as making sure it defines SSE2 and AVX2 itself if necessary. On my (Windows) system it doesn't compile at least the SSE2 stuff without this, so in SDL2 we must be getting that define from SDL itself.

[ankith26]

LGTM, thanks for the PR 🎉