What changed in EDuke32, build by build

Split MD3 model post-processing into common and Polymer parts and defer the latter if the game is started with a different renderer. This has the benefit of faster startup times and much less memory consumption in that case. When the user switches to Polymer later in-game, all processing is done in one run. I've observed crashes due to the music system if it takes too long, so the it may be unreliable. Also some calloc->malloc where the clearing is unnecessary.

Fix the build for everything else; it looks like the macro definition for

fix r1740 msvc build breakage

Revert the four Makefiles to have a base version in case I screw up again...

Fixes former crash in drawline16 and swaps drawpixel with the unsafe variant there; fixes potential sector[-1] access in sliding Star Trek doors; make CC overridable in Makefiles so that other toolchains or static analyzers like scan-build/clang can be used; with its help, fix some issues, some more severe than others.

Fix a bug where having two or more lighting passes on a multi-surface mdsprite would leak some GL state and disable blending for all but the first surface.

This commit doesn't add anything interesting for anyone not interested in tracing down completely absurd bugs.

Fix digitalnumberz spacing and saving/loading of screen size

remove obsolete CON compile warnings about affecting multiplayer sync, add -nodinput compatibility parameter to work around DirectInput crash issues on certain systems, fix http://forums.duke4.net/index.php?s=&showtopic=2961&view=findpost&p=60752

Set union ['] and difference [;] for sector selection in Mapster32; probably fix for crash on clipshape-init; replace tabs in doc/build*.txt with three spaces.

a.m32 corruption checker: also handle out-of-range picnum

Use posix_memalign in fork/exec midi player when not using nedmalloc. This fixes compilation in this case for me.

Works around a bug with newer NVIDIA GL drivers where creating a GL context

Fixes quite a few nasty bugs with the new clipping code, including the crash. Clip shapes are now loaded from _clipshape0.map through _clipshape9.map (no break like when loading tiles, just see if each is present). Mapster32: insert some safety code for sprites with sectnum<0. Because this is still a bad thing, have the m32script corruption checker catch this and run it periodically from EVENT_DRAW2DSCREEN. Also draw such sprites in 2D mode with a blood-red color, and make them draggable. Remove set[sector,wall,sprite] commands.

New sprite clipping for hitscan; crude debugging code for tracing down the crash that Commander is getting and I can reproduce only in real Windows with the synthesis build.

Fix crash. Reason was obsolete code.

SDL and menu joystick fixes by NY00123 (thanks!).

New CON command 'clipmovenoslide', doing the same as clipmove but stopping at collisions before sliding calculations. Fixes a problem with Mapster32 where names.h was tried to be read before setting up the search path.

New CON commands:

mapster32 fixes: aiming crash in polymer, more accurate aiming in polymost + calibration/replacement crosshair via "r_preview_mouseaim", 2 other crashes; fix "S" in tile selector, make read-only variable overridable with "script_expertmode", misc. beautifying.

msvc compile fix