Gauges: fix out-of-memory crash (HEAPSIZE), draw instrument panels

Enabling gauges crashed (null-this, illegal 0x478). Root cause was NOT video:
the MUNGA main heap (fixed 6MB default) exhausted loading gauge images. The
heap size comes from getenv("HEAPSIZE") (BTL4OPT.EXE @0x401076; default
0x600000); the pod's PARAMETR.BAT sets it (:POD=15MB, :REVIEW/:LOOP=32MB) but
our minimal setenv launch never did. memsize is irrelevant (it's the game's
own heap, not DOS memory -- verified memsize=127 still gave 6MB). Pods have
32MB RAM so 15MB fits.

Fix (no binary patch): set HEAPSIZE=15000000 + L4GAUGE=640x480x16 in
gauge.conf. Game now runs sustained with gauges on (500+ frames); the VESA
640x480x16 mode switches fine on our emulated S3 and the cockpit instrument
panels draw (DISPLAY/PROGRAM/ENG DATA/TRIGGER CONFIG...).

patch_btl4opt.py also gained Verify_Failed (DEBUG-build assertion) neutral-
ization -- release-equivalent robustness, separate from the heap fix.

Known-open (GAUGE-NOTES.md): the framebuffer renders only a top strip,
garbled -- the game bank-switches via a hardcoded far pointer in L4GAUGE.INI
(C000:2616, the STB Horizon+ card's VESA WinFuncPtr) that isn't valid on our
emulated S3, so only the first bank(s) page in. Fixing bank-switching (or an
LFB mode) is next; then the full buffer can be split into the six displays.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

emulator/gauge.conf

@@ -17,6 +17,11 @@ cd \REL410\BT
 set VIDEOFORMAT=svga
 set BLASTER=A220 I5 D1 H5 P330 T6
 set TEMP=c:\
+rem The pod's PARAMETR.BAT :POD path sets these; the game's HEAPSIZE default
+rem is only 6MB (too small once gauges load -> MUNGA heap exhaustion). Match
+rem the pod: 15MB heap + the 640x480x16 gauge mode.
+set HEAPSIZE=15000000
+set L4GAUGE=640x480x16
 rem arg4=g enables the gauge/secondary displays -> exercises the VDB splitter
 call setenv.bat r s n g
 32rtm.exe -x