docs(pdftract-5cqy): add xref stream parser verification note

The xref stream parser implementation was already complete in
crates/pdftract-core/src/parser/xref.rs. All acceptance criteria pass:

- Simple test /W [1 4 2] /Index [0 6]: 6 entries decoded correctly
- Type-2 compressed entries: route through ObjStm correctly
- Multi-subsection /Index [0 3 100 2]: produces correct entries
- Predictor support: FlateDecode + PNG predictor handled
- Zero-width field /W [1 4 0]: generation defaults to 0
- proptest: random byte sequences never panic
- INV-8 maintained: no production panics

All 11 xref stream tests pass.

Co-Authored-By: Claude Code <noreply@anthropic.com>

notes/pdftract-5cqy.md

@@ -0,0 +1,72 @@
+# pdftract-5cqy: Xref Stream Parser Implementation
+
+## Summary
+
+Implemented xref stream parser for PDF 1.5+ cross-reference streams with full support for:
+- `/W` field widths (type_w, obj_w, gen_w)
+- Type 0 (free), Type 1 (in-use), Type 2 (compressed in ObjStm) entries
+- `/Index` subsection boundaries with default `[0 /Size]`
+- Big-endian multi-byte field decoding
+- Zero-width field handling
+- FlateDecode decompression with PNG predictor support
+- Proper error handling and diagnostics (INV-8 compliant)
+
+## Implementation Location
+
+- File: `crates/pdftract-core/src/parser/xref.rs`
+- Function: `parse_xref_stream(source: &dyn PdfSource, stream_obj_offset: u64) -> XrefSection`
+- Lines: 1252-1569
+
+## Key Features
+
+1. **Indirect object parsing**: Uses Phase 1.2's `ObjectParser::parse_indirect_object()` to read the xref stream object
+2. **Stream decompression**: Uses Phase 1.5's `decode_stream()` for FlateDecode with predictor support
+3. **Field width handling**: Supports any `/W [type_w obj_w gen_w]` configuration including zero-width fields
+4. **Multi-subsection support**: Handles `/Index [first_1 count_1 first_2 count_2 ...]` arrays
+5. **Big-endian decoding**: `read_big_endian_field()` helper for 1-8 byte fields
+6. **Trailer dict extraction**: Copies relevant keys (Root, Info, ID, Encrypt, Prev) from stream dict
+
+## Test Results
+
+All xref stream tests pass:
+- `test_parse_xref_stream_simple`: PASS - /W [1 4 2] /Index [0 6] with 6 entries
+- `test_parse_xref_stream_multi_subsection`: PASS - /Index [0 3 100 2] produces correct entries
+- `test_parse_xref_stream_type2_compressed`: PASS - Type-2 entries route through ObjStm
+- `test_parse_xref_stream_field_width_zero_gen`: PASS - /W [1 4 0] (gen always 0)
+- `test_parse_xref_stream_with_predictor`: PASS - FlateDecode + PNG predictor
+- `test_parse_xref_stream_invalid_entry_type`: PASS - Unknown types emit diagnostics
+- `test_parse_xref_stream_missing_size`: PASS - Emits appropriate diagnostic
+- `test_parse_xref_stream_invalid_w_array`: PASS - Emits appropriate diagnostic
+- `proptest_parse_xref_stream_no_panic`: PASS - Random bytes never panic
+- `proptest_parse_xref_stream_random_offset_no_panic`: PASS - Random offsets never panic
+- `test_debug_xref_stream_parsing`: PASS - Debug helper test
+
+## INV-8 Compliance
+
+Verified: No `unwrap()`, `expect()`, or `panic!()` in production xref stream parsing code.
+- Line 206: `.unwrap_or(false)` is safe (handles poisoned lock gracefully)
+- All other `unwrap()`/`panic!` calls are in `#[cfg(test)]` modules (allowed per INV-8)
+
+## Acceptance Criteria Status
+
+| Criterion | Status | Notes |
+|-----------|--------|-------|
+| Simple test /W [1 4 2] /Index [0 6] | ✅ PASS | 6 entries decoded correctly |
+| Type-2 ObjStm routing | ✅ PASS | Compressed entries parse correctly |
+| Multi-subsection /Index [0 3 100 2] | ✅ PASS | Entries at 0,1,2,100,101 |
+| Predictor (FlateDecode + PNG) | ✅ PASS | Stream decoder handles transparently |
+| Field width /W [1 4 0] | ✅ PASS | Zero-width gen field defaults to 0 |
+| proptest random bytes | ✅ PASS | No panics on random input |
+| INV-8 maintained | ✅ PASS | No production panics |
+
+## Integration Points
+
+- **Phase 1.2**: Uses `ObjectParser::parse_indirect_object()` for reading the xref stream object
+- **Phase 1.5**: Uses `decode_stream()` for decompression with filter/predictor support
+- **Object resolution**: Type-2 entries return `XrefEntry::Compressed { obj_stm_nr, index }` for ObjStm resolver
+
+## References
+
+- Plan section: Phase 1.3 line 1089-1123 (xref streams)
+- PDF spec 7.5.8 (Cross-Reference Streams)
+- Bead: pdftract-5cqy