Apply localize minimum coverage before aggregation

[SuhasSrinivasan]

Fixes #651.

Summary

• Pass --min-coverage into localize record processing.
• Retain each bedMethyl record only when valid_coverage >= min_coverage, before offset aggregation.
• Add an exact end-to-end CLI regression covering below, equal-to, and above-threshold records.

Severity

Severity: High — scientific correctness

Rationale: The accepted option previously had no effect, silently allowing excluded low-coverage records to change localized scientific counts and percentages.

Root cause

EntryLocalize::run assigned and logged self.min_coverage but did not pass it into GenomeRegion::into_localized_mod_counts. The aggregation therefore folded every fetched bedMethyl row into n_valid and n_mod regardless of the requested threshold.

Implementation

Pass the minimum coverage into localized counting and filter each bedMethyl record before aggregation using the inclusive predicate valid_coverage >= min_coverage. The CLI regression uses two records that aggregate into the same output bin, with (valid coverage, modified count) values (1,1) and (23,2).

Preserved behavior

• Records equal to the threshold remain included.
• Aggregation, output ordering, and schema are unchanged after eligibility filtering.
• Default behavior is unchanged for records meeting the default threshold.

Non-goals

• Does not change localize anchor, window, or offset conventions.
• Does not define or implement --batch-size behavior.
• Does not change chart rendering or regional worker-error handling.

Behavior before and after

Case	Before	After	Expected oracle
--min-coverage 1	24 valid / 3 modified / 12.5%	Same	Both records meet the inclusive threshold
--min-coverage 3	24 valid / 3 modified / 12.5%	23 valid / 2 modified / 8.695652%	Coverage-1 record excluded before aggregation

Testing

Test environment

• Revision tested: 57c61c561a44addafe04f9b44dfd6de01cff5f95
• Tree tested and worktree state: e476b271f2bba441f9933c1894a16bf7b5d51096; clean at final verification
• Platform: macOS on Apple silicon
• Reference binary: installed modkit 0.6.4
• Exact Rust and tabix versions were not retained in the original test log.
• Dependency resolution: existing workspace manifests; no dependency or lockfile change in this PR

Test layer	Exact command, fixture, or matrix	Result and evidence
Core: parent-red regression	Final CLI regression applied to base 5cecc3fb3a9336068d9e3c68d5c08d678153dd2c	Failed as expected: threshold 3 still produced 24 / 3 / 12.5 instead of 23 / 2 / 8.695652
Core: focused regression	cargo test -p modkit --test test_localize test_localize_min_coverage_filters_before_aggregation -- --exact	Passed
Core: affected integration tests	cargo test -p modkit --test test_localize	2 passed, 0 failed
Core: applicable full workspace gate	cargo test --workspace --quiet -- --test-threads=1	Passed; aggregate count was not retained in the original log
Installed-versus-fixed comparison	Issue fixture at thresholds 1 and 3	Installed outputs are identical; fixed output changes only threshold 3 to the exact 23/2/8.695652 oracle
Core: formatting/diff checks	rustfmt --check --edition 2021 modkit-core/src/localise/subcommand.rs modkit-core/src/localise/util.rs modkit/tests/test_localize.rs; git diff --check	Passed

The serial workspace mode avoids an existing unrelated collision between pileup integration tests that share a temporary output filename.

Tests not performed

• No performance/RSS benchmark was run; the change adds one scalar comparison per fetched bedMethyl record.
• No large real-data regression was run; the two-record fixture distinguishes per-record filtering from an incorrect post-aggregation filter exactly.

Scientific validation

• Eligibility invariant: A record contributes if and only if valid_coverage >= min_coverage.
• Aggregation invariant: Filtering occurs before records sharing an output offset are combined.
• Count conservation: Threshold 1 sums (1,1)+(23,2) to (24,3); threshold 3 retains (23,2) only.
• Independent oracle: Direct inspection with tabix supplies the two source coverage/count pairs.

Output and compatibility

• Only rows below the requested threshold are removed from aggregates.
• TSV schema, offset ordering, and CLI syntax are unchanged.
• Threshold-boundary records remain included.

Reviewer guide

1. Review the two-record CLI fixture and its pre-aggregation oracle.
2. Review the threshold plumbing into GenomeRegion::into_localized_mod_counts.
3. Rerun cargo test -p modkit --test test_localize test_localize_min_coverage_filters_before_aggregation -- --exact.
4. Rerun cargo test -p modkit --test test_localize as the affected integration canary.

Checklist

• Linked issue contains reproducible observed and expected behavior.
• Change is limited to the linked issue's scope.
• Regression is red on the parent and green on this revision.
• All tests performed are listed with their results.
• Unrun applicable tests are disclosed.
• Scientific eligibility/count invariants and compatibility are checked.
• Changed-file formatting and diff hygiene pass.