Fix crash (0xC0000005) when a NULL city pointer is passed to City_has_improvement

[orcutt989]

Fixes a recurring 0xC0000005 crash (EIP 0x000002DF, a virtual call through a NULL vtable) triggered when patch_City_has_improvement() is invoked with a NULL City*.

Root cause

patch_City_has_improvement() unconditionally forwarded to the game's virtual City_has_improvement(). With a NULL this, the vtable dispatch executed call [NULL + 0x2DF], faulting on the read at 0x2DF.

Confirmed by disassembling the faulting wrapper in a full crash dump: the instruction sequence at the wrapper entry matched this function exactly (the share_wonders_in_hotseat branch, the this->Body.CivID reads, the p_human_player_bits mask, and the player-bits loop all line up instruction-for-instruction).

Fix

Added a guard at the top of the function returning false:

if (this == NULL)
    return false;

This is safe because all call sites consume the return value only in boolean context, and the code already NULL-guards this exact call elsewhere:

if ((city == NULL) || (! patch_City_has_improvement (city, __, wonder_improv_id, false)))

Notes

• Not config-dependent (reproduces with all-default config).
• Verified the compiled guard is present in the patched .c3xtxt section, and that injected_code.c compiles cleanly via the repo's TEST_INJECTED_CODE_COMPILE.bat.
• Changelog entry added.

[maxpetul]

Thanks for the contribution. I'd just like to know how you encountered this crash. What did you do in-game to trigger it? The changelog should refer to that and say something like "Fix possible crash when ___" instead of describing the changes made to the code, which are not meaningful for users.