perf(encoding): implement row-based match finder for fast/dfast levels

[polaz]

Problem

The donor enables a row-based match finder for greedy/lazy families and resolves it dynamically from parameters (see zstd_compress.c). Our encoder currently uses simple hash, dfast, and hash-chain paths only, without a row-based matcher backend.

That keeps low/mid compression levels behind donor on speed for common corpora.

Goal

Add a row-based match finder backend in Rust for Fastest/Default-oriented levels, with level-parameter routing similar to donor strategy selection.

Implementation plan

1. Add a RowMatch backend in zstd/src/encoding/match_generator.rs with dedicated table layout and probe logic.
2. Integrate backend selection into level parameter resolution so target levels can route to RowMatch.
3. Preserve current backends as fallback and for levels where RowMatch is not beneficial.
4. Add focused benchmarks for small and mid blocks where row match finder should improve throughput.
5. Validate ratio and interoperability against C zstd on the existing bench matrix.

Acceptance criteria

• RowMatch backend is selectable via level parameters and used in runtime for configured levels.
• Encoder output remains valid and decompressible by C zstd and our decoder.
• Throughput improves on target corpora at affected levels without large ratio regressions.
• Existing matcher tests keep passing; new backend-specific tests are added.

Dependencies

• Independent from perf: packed FSE Entry layout (4-byte entries + bulk table spreading) #56 and perf(decoding): SIMD HUF decode kernels with runtime CPU dispatch #66.
• Should be coordinated with feat: configurable compression parameters API #27 (configurable compression parameters) for future manual tuning.

Estimate

3d

[coderabbitai]

🔗 Related PRs

#34 - fix(encoding): implement default compression level [merged]
#39 - perf(encoding): align fastest matcher with zstd fast path [merged]
#41 - perf(encoding): interleave fastest hash fill insertion [merged]
#48 - feat(encoding): add Better compression level (zstd level 7, lazy2 strategy) [merged]
#53 - feat(encoding): add Best compression level (zstd level 11, btlazy2 strategy) [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!