Normalize `taxonomy` response shape across sequence-based and accession-based targets

[bencap]

Background

After #697, taxonomy is stored on TargetGene for both target types but surfaces differently in GET responses:

• Sequence-based: target_gene.target_sequence.taxonomy — preserved from the existing response shape via serialization shim
• Accession-based: target_gene.taxonomy — new field added by #CHILD2

This inconsistency is intentional in #CHILD2 to avoid a breaking response change, but it should not be a permanent state. Clients must handle two different locations for the same concept depending on which target type they are reading.

Proposed Changes

Expose taxonomy at target_gene.taxonomy for both target types. Remove taxonomy from the target_sequence response object, which then contains only sequence-type, label, and literal sequence data.

This is a breaking response change and should be coordinated and communicated with consumers.

Breaking Changes

	Change	Breaking?
TargetSequence response	Remove taxonomy field	response
TargetGene response	taxonomy present for both target types	additive for sequence-based

Migration Notes

No data migration required. This is a serialization-only change — taxonomy_id already lives on TargetGene after #697. Only the view model and any client code reading target_sequence.taxonomy needs updating.