Consolidate build_dedup_query() into bio-formats to eliminate duplicate query logic

[mwiewior]

Problem

The transcript dedup + HGNC propagation query is duplicated in two repos:

1. datafusion-bio-formats — storable_to_parquet.rs → build_dedup_query()
2. datafusion-bio-functions — cache_builder.rs → build_query() / build_query_multi_chrom()

When the HGNC propagation fix (biodatageeks/datafusion-bio-formats#162) was added to bio-formats, the identical query in bio-functions was missed, causing #105 to persist until 33d700f ported the fix manually.

Proposed fix

Export build_dedup_query() from datafusion-bio-format-ensembl-cache as a public library function. Then cache_builder.rs in bio-functions replaces its local build_query() with a call to the shared function.

The parallel I/O pipeline, writer tuning, and chrom-splitting in cache_builder.rs stay untouched — only the SQL query construction moves to bio-formats.

Context

• fix: replicate VEP merge_features HGNC_ID propagation in cache builder datafusion-bio-formats#162 — original HGNC propagation fix
• HGNC_ID gap: cache doesn't replicate VEP's merge_features symbol-to-HGNC propagation #105 — HGNC_ID gap issue
• 33d700f — manual port of the fix to cache_builder