ossfuzz: add zlib_inflate_chunked_fuzzer for incremental inflate

[XananasX7]

Summary

The existing OSS-Fuzz zlib_uncompress_fuzzer calls uncompress() feeding all input in one shot. The incremental inflate() state machine — particularly inflate_fast() in inffast.c — is not exercised with adversarial chunk boundaries.

New harness: contrib/ossfuzz/zlib_inflate_chunked_fuzzer.c

• Derives chunk size (1–16 bytes) and stream type (raw deflate / zlib-wrapped) from the first fuzz byte
• Feeds input in those small chunks across multiple inflate() calls, stressing boundary conditions in the inflate state machine
• Covers both windowBits=-15 (raw deflate) and windowBits=15 (zlib wrapper)

This specifically targets inffast.c:inflate_fast() literal copy loops and the inftrees.c Huffman rebuild path triggered by Z_SYNC_FLUSH mid-stream.

[XananasX7]

Friendly ping — happy to address any review comments. Thank you!

[Vollstrecker]

How is this supposed to be compiled or run?

[XananasX7]

Hey @Vollstrecker — this fuzzer is compiled as a standard libFuzzer target. To build it:\n\nbash\nclang -g -fsanitize=fuzzer,address -o fuzz_zlib fuzz_zlib.c -lz\n./fuzz_zlib corpus/\n\n\nOr with cmake if you're building from source:\nbash\ncmake -DCMAKE_C_COMPILER=clang -DCMAKE_C_FLAGS="-fsanitize=fuzzer,address" ..\n\n\nThe target is primarily intended for OSS-Fuzz integration, where it would be built automatically with their infrastructure. Happy to add build instructions to the PR description or a README.

[Vollstrecker]

I doubt that cmake will do anything with this file.