Made some efforts to improve efficiency of new code, hard to tell if I've made it better or worse though honestly

R_IsEmptyLine is now a thing too btw

More progress, NEWCLIP added to doomdef.h, sadly it actually all lags the game so I've disabled it for now

Other notes:
* on second thought I'll keep the hw_clip functions' gld prefixes rather than HWR, not like it matters either way
* despite the extra lag it does fix the issues with translucent walls and such when displayed at different vertical angles, such as with the GFZ1 waterfall

Other notes:
* Renamed all new functions to have HWR_ prefix instead of gld_, for consistency
* HWR_FrustrumSetup and HWR_SphereInFrustum are disabled and require HAVE_SPHEREFRUSTRUM. This is because 1) SRB2CB did not need the code, so presumably neither will we, and 2) there are some OpenGL API functions used there that due to our way of using OpenGL we don't use outside of r_opengl.c, which makes dealing with HWR_FrustrumSetup complicated in theory
* The new clipping functions are not added to OpenGL's "main" rendering code itself just yet, they're just available to use now once hw_clip.h is included

* Move the misc types in the file to bottom, so that ATTRPACK at least is usable for RGBA_t
* Include endian.h, so that UINT2RGBA can be defined correctly for big endian builds
* Add more comments to make clear the main sections of the file

Remove FUNCMATH from functions with a void return value or args, or examine variables other than their args (which could be modified)

D_PostEvent_end is only used by Allegro (used by the DOS port) to help timers work, so check for PC_DOS in preprocessor code. Also remove FUNCMATH from said function.

P_NullPrecipThinker no longer should have FUNCMATH (though I'm not sure if it should have had it in the first place anyway)

Make sure handles of files opened by W_LoadWadFile are closed if we abort loading the files for whatever reason.

ESPECIALLY if the file is already loaded in SRB2, that's just silly.

now that V_FLIP's effects are always supported in V_DrawFixedPatch, there's no need for the "flip" variable anymore

HWR_DrawFixedPatch: Add V_OFFSET support for V_FLIP in hardware code as well, also add missing SHORTs for gpatch fields

Make V_OFFSET no thing without V_NOSCALESTART, to remove any potential glitches with using it on its own

Removed all remaining traces of VID_X11 code in hw_drv.h and hw_data.h, the macro is no longer used by Linux etc versions of SRB2.

Unlike the rest of the commits in this branch (as of writing), I didn't make this commit between 1 and 2 years ago, I made it right now ;)

Remove pointless drawtextured variable and redundant fake planes checks (R_FakeFlat would already have made gr_frontsector/backsector something else if they were)

Added missing checks in HWR_AddLine from the software version, move v** vars to the top since R_PointToAngle calls use the same values anyway

this fixes a crash in (old) GFZ2 at the ramp as a result of creating pv1/pv2. This probably means before pv1/pv2 there could have been some silly typecasting from vertex_t to polyvertex_t to get fixed vertex coords and such...

Yes I know they're void * in r_defs.h's seg_t definition, it's quicker than trying to figure out if including hardware/hw_glob.h is a good idea or not

I also optimised those two functions while I was there (why keep a "floating" variable when setting it to false guarantees the functions return false?)

gr_correcttricks fix: don't check if top/bottom textures are missing for sloped sectors, just ignore and cancel the hack

This fixes GFZ2's mysterious flying flats at the ramp to the big room

Add MD2_INDENT and MD2_VERSION so we can cleanly check that it's a valid MD2 from magic number/version

Use temporary buffer with a max size of 255 bytes instead of having Microsoft's FormatMessageA alloc one for us. Also, provide a fallback message in case no message was available for some reason