[P2] PNG and UF2 LZMA decompression has no memory or output-size bounds #13

The embedded-source decoder calls lzma.decompress without a decoder memory limit or a bound on decompressed output. The resulting data is then copied into UTF-16 and parsed as JSON, further increasing peak memory usage.

Relevant code: embedded_source.py, lines 79–82.

The UF2 image-size limit bounds the compressed input image, not the expanded source. ZIP's uncompressed-size limit does not apply to PNG/UF2 imports.

A bounded review probe constructed a valid 1,458-byte MakeCode project PNG that successfully extracted 16 MiB of source. Larger compressed payloads can exhaust a build worker's memory despite the other import limits.

Expected behavior: bound both the LZMA decoder's dictionary allocation and the total expanded output before decoding/copying/parsing the complete project, and report an import error when the limits are exceeded. A decoder memory limit alone does not cap the expanded output.

Found while reviewing PR #7, toolchain at ce12c48.