feat: taxonomy info change and assembly manifest update

[emilylm]

📌 Summary

This PR improves the taxonomy_info and assembly intent endpoints with several related enhancements. On the taxonomy side, it adds bulk upsert and NCBI refresh endpoints so existing records can be updated without re-importing from scratch. It also removes redundant columns. On the assembly side, the create manifest endpoint now accepts multiple Hi-C specimen sample IDs, allowing assemblies that draw Hi-C reads from more than one specimen to be correctly represented in the generated manifest. Some manifest fields were also updated to match the intended schema.

---

🔄 Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 💥 Breaking change (fix or feature that would cause existing functionality to change)
• 🧹 Refactor (code change that neither fixes a bug nor adds a feature)
• 📚 Documentation update
• ⚙️ CI / build / tooling change

---

🧩 Key Changes

• POST /taxonomy-info/bulk-upsert — new endpoint that inserts or updates taxonomy info rows and re-fetches NCBI data, mirroring the behaviour of the single-record create endpoint
• POST /taxonomy-info/bulk-ncbi-refresh — new endpoint that re-syncs only the ncbi_* columns for a supplied list of existing taxon_ids, leaving upstream fields unchanged
• Multiple Hi-C specimens in assembly intent — AssemblyIntent.hic_specimen_sample_id (single UUID) replaced by hic_specimen_sample_ids (list of UUIDs); the manifest helper's routing check is now a set membership test, requiring minimal logic change
• Taxonomy schema clean-up — removed redundant/outdated columns (defined_class, mito_ref, busco_dataset_name) and backfilled mitohifi_reference_species from mito_ref
• Manifest field updates — added assembly_id to the generated manifest; removed PacBio file extension validation to be consistent with ONT and Hi-C handling
• Migration 0003 — adds nullable hic_specimen_sample_ids JSONB column to assembly; the existing hic_specimen_sample_id FK column is retained and set to the first entry for backwards compatibility; removes defined_class (redundant with ncbi_class), mito_ref (redundant with mitohifi_reference_species), and
  busco_dataset_name (replaced by busco_odb10_dataset_name and busco_odb12_dataset_name).

---

✅ Checklist

• Tests are passing locally / in CI
• New tests have been added for new or changed functionality
• Self-review completed
• Documentation updated (if needed)
• No sensitive data or secrets included

---

🔍 Review Notes

• The four pre-existing test failures in test_endpoints_assemblies.py (manifest response shape, reads vs read_files key) were already failing on main before this branch — they are not regressions introduced here
• hic_specimen_sample_ids has no DB-level FK constraint (stored as JSONB); referential integrity is enforced at the API layer via _validate_specimen_sample before any write. The long_read_specimen_sample_id remains the primary identifier for versioning and assembly identity
• The bulk-ncbi-refresh endpoint skips rows that don't yet have a taxonomy_info record — callers should use bulk-import or bulk-upsert first if creating new rows is also needed

---

📎 Related Issues / Tickets

• Closes #
• Related to #