perf(encoder): strategy-aware literal compression gates G4+G5 (skipLitCheck + minGain)

[polaz]

Context

Follow-up to #126 (G3 whole-block bail-out, merged). Donor
ZSTD_compressLiterals has two more strategy-dependent gates we don't
yet honour:

G4 — skipLitCheck: when the strategy emits a synthetic
single-literal RLE-like block, donor skips the HUF compression attempt
entirely (zstd_compress_literals.c huf_compress_skipLitCheck path).

G5 — minGain: even after computing the HUF-compressed size, donor
requires the gain to exceed a strategy-specific minimum threshold
before emitting compressed literals (HUF_compress_internal → minGain
formula varies per strategy: fast/dfast use a tighter threshold than
greedy/lazy/btultra).

Our path emits compressed literals whenever HUF reduces size by any
amount — produces sub-optimal frames vs donor on Fast/dfast levels.

Acceptance criteria

• Port skipLitCheck short-circuit for synthetic RLE-like blocks
• Port strategy-aware minGain formula matching donor per-strategy
  thresholds (cite donor file:line for each strategy)
• compare_ffi REPORT: no ratio regression; Fast/dfast bytes within
  ±0.5% of donor on decodecorpus-* corpus
• Unit tests for the gate decisions per strategy
• No speed regression on encode hot path

Files involved

• zstd/src/encoding/blocks/compressed.rs (where G3 currently lives)
• zstd/src/huff0/huff0_encoder.rs (size estimator if needed)

References

• Part of perf(decode): HUF 4-stream rewrite — sentinel-bit refill + 5-symbol unroll + batch RELOAD + const-generic kernel #199-adjacent encoder ratio sweep
• Donor: zstd_compress_literals.c, HUF_compress_internal

[coderabbitai]

🔗 Related PRs

#125 - perf(encoding): #124 Phase 4 — HC speculative tail check + MatcherStorage enum dispatch [merged]
#179 - perf(encoder): donor-parity greedy parse at L4 — ratio + speed win [merged]
#181 - perf(encoder): G3 — whole-block bail-out before partition split [merged]
#182 - feat(encoder): strategy-aware literal gates (G4 + G5) [merged]
#201 - perf(decode): port donor HUF 4-stream burst with sentinel-bit ctz [closed]

---

📝 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!