Remove dead genebody coverage code

[ewels]

Summary

• Removes the legacy GenebodyCoverageAccum / TranscriptPositionMap code that was computed in the counting hot loop but whose output (CountResult.genebody) was never read or written anywhere
• This code was fully superseded by the qualimap module (PR Add Qualimap RNA-Seq QC module #19) — the field was marked #[allow(dead_code)] and docs already called it "legacy"
• Eliminates wasted per-read computation (O(aligned_blocks × exons) per assigned read) and ~540 lines of dead code

Test plan

• cargo build — compiles cleanly
• cargo test — all 184 tests pass (172 unit + 12 integration)
• cargo clippy — no new warnings (4 pre-existing doc indentation warnings in unrelated code)
• Verify output files are identical before/after on a real dataset

🤖 Generated with Claude Code

[netlify]

✅ Deploy Preview for rustqc canceled.

Name	Link
🔨 Latest commit	6bd8064
🔍 Latest deploy log	https://app.netlify.com/projects/rustqc/deploys/69a8a11e28adc30008fae458

[ewels]

The follow-up commits (622319e, 6bd8064) go beyond the original genebody removal and clean up additional dead code and issues found across the codebase:

Removed unused fields

• BamStatResult: removed mapq_distribution and primary_paired (populated by the accumulator but never read by any output writer)
• QualimapResult: removed fragment_count, transcript_coverage, and merged_gene_coverage (computed but never used in reports)
• QualimapIndex: removed num_genes (only used in a log message, replaced by gene_transcript_ranges.len())
• MergedGeneModel: removed gene_idx and exonic_length (set during construction but never read; exonic_length is now computed on-the-fly where needed)
• ExonMeta: removed transcript_idx (stored in COITree metadata but never accessed)
• TranscriptCoverage::iter(): removed (only caller was the now-removed merged_gene_coverage code)
• QualimapAccum::into_result(): removed unused index parameter (was only needed for removed transcript_coverage building)

Fixed incorrect #[allow(dead_code)] annotations

• Several fields had #[allow(dead_code)] with comments like "kept for API completeness" when they were genuinely unused — removed the fields instead of keeping the suppression
• ExonMeta and TranscriptSampling (tin.rs): kept #[allow(dead_code)] but updated comments to accurately explain why (COITree storage / values used during construction)

Fixed clippy issues

• Removed orphaned doc comment in counting.rs (from a previously-deleted function) that caused doc_lazy_continuation errors
• Fixed .clone() on Copy types in index.rs tests
• Fixed unnecessary as u64 cast in idxstats.rs
• Fixed &mut vec![...] → &mut [...] in output.rs tests

Regression testing

Built and ran both main and this branch on benchmark/input/small/test.bam — all QC output data is identical (only differences are output paths and timestamps).