fix(mitmproxy): decompress response bodies based on content-encoding

[Arkptz]

Root cause

mitmproxy v21 stores response bodies as raw wire bytes — still compressed per the content-encoding header. Our parser read them directly and passed compressed bytes to parse_body(), which tried serde_json::from_slice() on brotli/gzip data and silently returned None. Result: zero schemas emitted for any compressed response.

Fix

Added decompress_body(body: &[u8], encoding: Option<&str>) -> Vec<u8> helper that handles br (brotli), gzip, and deflate. On decompression failure it logs a warning and falls back to raw bytes for backward compatibility.

Applied to both request and response body extraction in parse_flow().

Test evidence

• 5 new unit tests: brotli round-trip, gzip round-trip, deflate round-trip, invalid data fallback, no-encoding passthrough
• cargo test: 240 passed, 0 failed
• Real MEXC capture: grep -c "schema:" /tmp/fixed-spec.yaml → 2 (was 0), success field found in schema properties

Files changed

• Cargo.toml: added brotli = "7" and flate2 = "1"
• src/mitmproxy_reader.rs: decompress_body helper + applied at request/response extraction points