CnvTranslator: c. HGVS inputs should be converted to n. before building SequenceLocation

Summary

CnvTranslator._from_hgvs currently builds a SequenceLocation directly from sv.posedit.pos.{start,end}.base for any supported HGVS input, including CDS-relative (c.) expressions. Per the VRS 2.0 specification, SequenceLocation.start/end are indices into the sequence identified by refgetAccession in that sequence's native frame — so when the refget accession points at a transcript (e.g. NM_*), the coordinates need to be transcript-relative (n.), not CDS-relative.

From VRS 2.0 — Implied Sequence Coordinates:

The Sequence Location class refers to coordinates on a SequenceReference; if that sequence represents a coding transcript, then the coordinates refer to the coding transcript, and not a chromosome sequence to which it aligns.

AlleleTranslator's extract_allele_values already handles this with a c_to_n step in HgvsTools. CnvTranslator does not, so a c. CNV input yields a SequenceLocation whose coordinates land in the 5' UTR instead of the intended CDS region — and whose ga4gh:SL.* digest differs from the correct value.

Example

For NM_001331029.1:c.100_200del (CDS positions 100–200 of a transcript with a 150-base 5' UTR):

Fix

Mirror the Allele path: after the existing is_intronic guard and derive_refget_accession lookup, run c_to_n when sv.type == "c". Regression test added using NM_001331029.1:c.100_200del against the expected transcript-relative output.

All existing g. CNV tests are unaffected (sv.type != "c", so the new branch doesn't fire). Consumers passing c. HGVS to CnvTranslator will see their output SequenceLocation coordinates and digests update to the transcript-relative values required by the spec.