CORRECTION: 0x4334d0 was NOT EncodeField (it's a CUtlVector realloc helper)

Disassembly of 0x4334d0 proved it's a memory/realloc helper cluster (IMemAlloc Alloc/
Realloc/GetSize vtable calls), not the field encoder — the earlier Ghidra decomps were
misattributed. Located the real reference: the 'EncodeField encoder wrote %d bits' string
(vaddr 0x15d580) is referenced at 0x3357fb, inside the function at entry 0x3356dd (Ghidra
mis-names it __numpunct_fill_cache). Updated the gamedata sig + module sig to that function.
CANDIDATE — static RE here is unreliable (wrong twice); the real per-tick encoder must be
confirmed dynamically (gdb + real client). Documented.

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