Introduce a mapping table at the target level

[bencap]

Introduce a mapping table that sits alongside the target record linked by FK and stores three categories of data:

• Alignment provenance: tool name and version, parameters used, reference accession aligned against, alignment level (protein/nucleotide), timestamp
• Alignment QC: percent identity, mismatch count and positions, gap count and positions, alignment score, next-best alignment score, alignment string
• Annotation QC: total variant count, count mapped cleanly, count mapped with warnings, count failed, failure mode breakdown

Mapping outputs (gene name, mapped reference accession, mapped coordinates) should stay on the target record for convenience. The mapping table is an auditing and QC layer, not a replacement for those fields. This table is a prerequisite for most other subtasks in this issue category.

This table will also let us de-duplicate some of the provenance information stored alongside mapped variants.

[bencap]

We should consider also persisting alignment files to S3.