fix(bam): emit SAM-spec NM:i: tag distinct from STAR-internal nM:i:

[pinin4fjords]

Summary

--outSAMattributes NM was routed to the existing nM:i: writer (substitutions only). The SAM-spec NM tag is the edit distance to the reference: substitutions + inserted bases + deleted bases. STAR and SAM-conformant tools (samtools stats, Picard, MultiQC) read NM:i: and were getting nothing back.

BAM	NM:i: records	nM:i: records
rustar (pre-fix)	0	88 592
STAR	88 072	0

On records with indels, the per-record values also differed (rustar nM:i:0 where STAR's NM:i:2 reflected the deleted bases).

Fix

Two changes:

1. In the attribute consumer, treat NM and nM as distinct tokens - request for one no longer silently enables the other.
2. When NM is requested, compute SAM-spec edit distance: n_mismatch + sum(I-op lengths) + sum(D-op lengths) from the existing Transcript::cigar and Transcript::n_mismatch. Intron N skips are excluded per SAM v1 section 1.4.

Both tags can now be emitted together when both are in --outSAMattributes.

Test plan

• New unit test asserts an 80M2D21M record with one mismatch produces NM:i:3 and nM:i:1
• Existing tests pass
• cargo build
• cargo clippy --lib -- -D warnings
• cargo fmt --check

After this fix, samtools view <bam> | grep -c 'NM:i:' matches the total record count and samtools stats error_rate reports a non-zero edit distance.

Fixes #29

[pinin4fjords]

Verified end-to-end on macOS/aarch64 against the rebuilt fix branch.

$ samtools view <bam> | grep -c 'NM:i:'    -> 88496
$ samtools view <bam> | grep -c 'nM:i:'    -> 0    (only NM was in --outSAMattributes)

Per-record NM values on indel-containing records match STAR's exactly:

CIGAR	n_mismatch	rustar NM:i:	STAR NM:i: (same read)
80M2D21M	0	2	2
43M1D58M	0	1	1
18M1I82M	0	1	1
77M1D24M	1	2	2
77M1I23M	0	1	1

samtools stats after the fix:

SN  mismatches:  77025  # from NM fields
SN  error rate:  9.074168e-03  # mismatches / bases mapped (cigar)

Pre-fix the same run produced 0 NM:i: tags and samtools stats reported mismatches: 0 / error rate: 0. LGTM.