fuzzers: add opj_compress_fuzzer for J2K/JP2 encode path

[XananasX7]

Summary

The existing OSS-Fuzz harnesses cover only decompression:

• opj_decompress_fuzzer_J2K – .j2k / .jph codestream decode
• opj_decompress_fuzzer_JP2 – .jp2 JP2 container decode

The encode path (src/lib/openjp2/j2k.c, jp2.c, tcd.c) has no OSS-Fuzz coverage.

New harness: tests/fuzzers/opj_compress_fuzzer.cpp

Derives image parameters (width, height, components, lossy flag, J2K vs JP2) from fuzz input, builds an opj_image_t with fuzz-derived pixel data, then runs a full encode via opj_setup_encoder → opj_start_compress → opj_encode → opj_end_compress, writing output to an in-memory write-stream using custom callbacks (matching the pattern of existing decompress fuzzers).

Covers: DWT forward transform, tier-1/tier-2 encoding, rate-distortion optimisation, JP2 box serialisation in jp2.c.

[XananasX7]

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

[rouault]

formatting fixes to address: https://github.com/uclouvain/openjpeg/actions/runs/27315401875/job/82500689891?pr=1645