replace: fix empty-index header skip, dead match tracking, port streaming parallel write

[jqnatividad]

Summary

Code review of src/cmd/replace.rs against the recent search/searchset parallel refactor (#3776) surfaced one observable bug, dead code, and an efficiency gap. This PR addresses all of them.

• Bug fix: parallel_replace skipped headers and ignored --not-one when the indexed file had zero data rows; sequential_replace always wrote headers. Both paths now emit headers up-front and route through handle_replace_results so --not-one semantics match.
• Bug fix: worker threads no longer .unwrap() on indexed() / seek() / byte_records() errors — CliResult is propagated through a closure, matching search.rs:582-810.
• Cleanup: drop the dead had_match field + _rows_with_matches_ctr counter from the parallel path (the sequential progress message now derives it from match_count > 0); make parallel_replace take &Regex (matching sequential_replace); drop a redundant record = processed_record; reassignment, a stale // self.clone(), comment, the #[allow(dead_code)] on Args, and the unqualified use crossbeam_channel;.
• Streaming parallel write: replace the whole-file Vec<Vec<ReplaceResult>> buffer + sort with the BTreeMap<usize, ChunkOutput> + next_chunk cursor pattern from search.rs. Peak memory is now ~njobs × chunk_size records instead of the entire file. Per-row row_number: u64 is gone — chunks carry chunk_index for ordering; within-chunk order is preserved by push order.

Test plan

• cargo build --locked --bin qsv -F all_features
• cargo test --locked -F all_features replace — 31/31 passing, including existing replace_indexed_parallel (multi-chunk ordering with --jobs 2) and the new replace_indexed_empty regression
• cargo clippy --locked -F all_features --bin qsv -- -D warnings — clean

🤖 Generated with Claude Code

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 7 complexity

Metric	Results
Complexity	7

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[Copilot]

Pull request overview

This PR improves qsv replace’s indexed/parallel execution path to match the sequential path’s semantics (especially for empty indexed inputs) and to reduce peak memory usage by streaming parallel chunk outputs in input order.

Changes:

• Fix parallel replace behavior for indexed files with zero data rows: headers are emitted and --not-one/exit-code semantics match sequential replace.
• Propagate worker I/O/index/CSV errors via CliResult instead of panicking, and remove dead match-tracking state in the parallel path.
• Switch parallel output from “buffer everything then sort” to a streaming BTreeMap + next_chunk ordering pattern (ported from search).

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/cmd/replace.rs	Aligns parallel replace semantics with sequential, removes dead tracking, propagates errors, and streams chunk outputs in order to bound memory.
tests/test_replace.rs	Adds a regression test covering the empty-indexed-file behavior (headers + --not-one semantics) under parallel mode.