enhancement: diff parser hardening + smart file splitting for reviews #195

Summary

The diff parser (src/engine/diff_parser.rs) and bundling engine (src/engine/bundling/) need hardening to handle edge cases encountered in real-world usage across 13+ repos. This is a prerequisite for #188 (auto-chunking large diffs).

Current State

Known Issues

1. Diff Parser Edge Cases

• No graceful handling of malformed diffs — if git output is incomplete or truncated, parser silently drops data
• Binary file detection is prefix-only (^Binary files) — may miss some binary markers
• No validation of hunk line counts — parsed old_count/new_count not verified against actual lines collected
• Regex on every line — performance concern for large diffs (thousands of lines)

2. Split/Chunking Gaps

• No file-level splitting — large single-file diffs cannot be split into hunk groups
• Bundling only groups files — does not split oversized groups
• No progress reporting during multi-chunk reviews
• Cross-file context lost when splitting (related changes in different files get separated)

Proposed Improvements

Phase 1: Diff Parser Hardening

• Add validation: verify hunk line counts match collected lines
• Handle truncated/incomplete diffs gracefully (warn + partial parse)
• Cache compiled regexes (already using LazyLock ✓)
• Add structured error type instead of silent failures

Phase 2: Smart Splitting

• Split oversized groups by hunk boundary (not mid-hunk)
• Preserve directory proximity when splitting — keep src/config/*.rs together
• Track cross-file references across splits for context continuity
• Add --split-strategy flag: hunk | file | directory

Phase 3: Integration with Auto-Chunking (#188)

• Wire smart splitting into the chunker module from enhancement: auto-chunking for large diffs #188
• Progress output: Reviewing chunk 3/10 (src/engine/*)…
• Dedup findings across chunks by (file, line, message) tuple

Target

• v0.5.0 (new functionality, no breaking changes)
• Depends on: none (can start independently)
• Related: enhancement: auto-chunking for large diffs #188 (auto-chunking), fix: handle optional capture groups in hunk header regex #167 (hunk header regex fix)