db/seg: collapse compression workers=1 case into batched workers flow

[sudeepdino008]

Summary

• collapse case of workers>1 and workers=1 in single flow
• this is a simplification; W=1 case was specially handled, removing that and merging with other flow leads to 1% slower merge file creation, which I consider acceptable.

Benchmark

Re-compressing the word stream of a real mainnet storage .kv (527 MB, 17.6M words / 8.8M keys, key compressed + value not, as CompressKeys does), 2 iterations x 2 rounds:

	base (#21997)	this PR
cover phase, W=1	10.02 s	10.29 s
cover phase, W=4 (control)	3.11 s	3.07 s
end-to-end, W=1	34.30 s	34.53 s

Workers > 1 is unchanged code, so those rows measure the noise floor (±1.5%).

A synthetic corpus (1M sorted 52-byte storage-like keys) resolves the Workers=1 cost more tightly — 10 iterations x 3 alternating rounds, medians varying under 1 ms within each version:

	base (#21997)	this PR
cover phase, W=1	322.0 / 321.8 / 322.5 ms	329.2 / 331.7 / 330.1 ms
end-to-end, W=1	1432 / 1431 / 1439 ms	1435 / 1442 / 1448 ms

So ~2.6% on the cover phase at Workers=1, under 1% end-to-end — the two extra CompressionWord copies plus the queue push/pop. Dictionary learning dominates a real merge, which is why the end-to-end cost stays inside the noise.

Notes

• Pure refactor, no behavior change, so no new red-green cycle — the existing tests plus the golden checksums are the safety net.

[Copilot]

Pull request overview

This PR refactors db/seg compression so the Workers=1 cover/compression path uses the same batched-worker pipeline as the Workers>1 path, removing the prior special-cased single-worker implementation. This simplifies the code while accepting a small Workers=1 performance regression, and updates tests to pin byte-identical output to pre-batching checksums across worker counts.

Changes:

• Remove the special-case single-worker cover/compression path and always run through the batched worker pipeline.
• Adjust cover batching sizing logic to apply uniformly, and unify queue/drain logic for all worker counts.
• Update parallel compression tests to validate round-trip correctness and stable on-disk encoding via fixed CRC32 checksums across worker counts.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
db/seg/parallel_compress.go	Collapses Workers=1 and Workers>1 compression flows into a single batched worker pipeline.
db/seg/compress_parallel_test.go	Updates helpers/tests to assert stable output encoding via fixed checksums across worker counts.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.