Correct BND input triggers a misleading WARN about missing END/SVLEN #497

Supplying a well-formed inter-chromosomal breakend via input_vcf logs:

WARN  Symbolic SV at 1-based POS 600 has no END/SVLEN — skipping coverage modulation

The behaviour is correct — a breakend is a point (VCF 4.2 §5.4: END == POS, no SVLEN) and carries no copy-number change, so skipping coverage modulation is exactly right. But at WARN level, phrased as something missing, it reads as a defect in the user's input.

This matters more now that input_vcf is the recommended path for modelling a specific rearrangement (de novo SVs are opt-in and not recommended for human germline). A researcher supplying a known translocation gets a warning implying their VCF is malformed, when it is spec-perfect.

Suggested fix: for SVTYPE=BND specifically, either drop to DEBUG or reword to state the expectation, e.g. "breakend at POS 600 — no coverage modulation applied (a breakend is a point event, as expected)". Keep the current WARN for genuinely under-specified symbolic SVs that are NOT breakends, since a <DEL> with no END/SVLEN really is malformed.

Repro:

# two-record reciprocal BND pair across contigs, sv_rate_scale: 0.0, input_vcf set
eidolon gen-reads -c cfg.yml