extmod/modframebuf.c: pixel index arithmetic overflows on 16-bit int targets (out-of-bounds writes)

The pixel index expressions in extmod/modframebuf.c — fourteen of them, shaped like x + y * fb->stride — are computed in (16-bit-promotable) integer arithmetic. With 16-bit int, a 256-wide buffer overflows the index from y ≥ 128 and reads/writes land out of bounds. On our port this corrupted the heap every time text was drawn on the lower half of a 256×192 framebuffer.

Environment: MicroPython v1.28.0, custom 65816 (SNES) port, Calypsi toolchain (16-bit int). Any 16-bit-int build is affected. Our fix (cast the operands to size_t in all 14 index expressions) is in patches/0001-compiler-workarounds.patch: https://github.com/FabianKuebler/micropython-snes (DECISIONS.md, "modframebuf 16-bit overflow" entry).

Analysis and drafting done with AI assistance; every finding verified on target/emulator.

Start in extmod/modframebuf.c and inspect the fourteen pixel index expressions shaped like x + y * fb->stride. Reproduce the issue on a 16-bit-int target or emulator, then verify all affected expressions use the intended size-safe arithmetic and no out-of-bounds writes occur for the 256×192 framebuffer. The proposed patches/0001-compiler-workarounds.patch and its DECISIONS.md entry provide additional context.