use ATTRINLINE rather than inline to match some of the other functions in this file, though I'm told modern compilers ignore inlining info as they actually decide themselves now

(though I could pretend we still support the older ones lol)

Turn the fixed-float convert macros into inline functions, like those in GZDoom. The old macros get to be just the same as calling these inline functions.

Updates and fixes for the ZB config

See merge request !859

Skip empty gametypes in level select menu

See merge request STJr/SRB2!819

Remove the multi-file ZB configs. We're stick with the single-file config for ZB and doing all the UDMF stuff in UDB

NiGHTS Mode menu fixes

See merge request !841

Fix division-by-0 crash from 100% gamepad deadzones, and lack of input from 0% deadzones

See merge request !838

Actually use the "SCREENSHOT OPTIONS" header for its menu.

See merge request !832

Skins refactor

See merge request STJr/SRB2!843

Fix NiGHTS Attack crash after render mode change

See merge request !837

Fix HWR_DrawFlippedColumnInCache ignoring chroma key TF

See merge request !826

Replace NextTexAvail with glGenTextures

See merge request !828

Fix ASTBlendPixel blending with nothing

See merge request !827

Archive sprite2 Hotfix

See merge request !845

Fix desyncs never recovering if node and player mismatch

See merge request !842

One line fix: don't assume 0 (aka SPR2_STND) is the default value for sprite2, but rather what the state sets for it

This fixes some issues with a custom character tested during netplay, which did not have SPR2_WAIT sprites and therefore fell back to SPR2_STND sprites. Unfortunately, the fact they used SPR2_STND instead meant the sprite2 was not synced at all!

(oddly enough, this actually revealed some secret file dependencies previously included via r_things.h! I also needed to include d_player.h in r_skins.h itself it seems)

Set the "spritename" variable within R_AddSingleSpriteDef rather than externally to it, since R_InstallSpriteLump is the only place it gets used anyway

This way we don't have to make it some extern in order for r_skins.c to use it

It makes more sense for 100% deadzone to just make it so that
you have to push the axis all the way to trigger it,
rather than 100% deadzone resulting in no axis input
ever happening... So, let's make it be the former way instead

Fix division-by-0 crash with gamepad deadzones
The problem was that it checked if A was more than B,
then lowered A to a max value, then subtracted B from A,
then divided something by that, without checking if A minus B was 0,
allowing division by 0 if B was the same as that max value

This fixes that by making sure that A is less than the max value

Fix -OGLlib

See merge request STJr/SRB2!794