finding.tags.add() per finding calls tagulous's add() which does:
  - reload() → SELECT current tags (1 query)
  - _ensure_tags_in_db() → get_or_create per tag (T queries)
  - super().add() → INSERT through-table rows (1 query)
  - tag.increment() → UPDATE count per tag (T queries)

For N findings with T parser-supplied tags: O(N·T) queries.

Replace with bulk_apply_parser_tags() in tag_utils, which groups
findings by tag name and calls bulk_add_tags_to_instances() once per
unique tag: O(unique_tags) queries regardless of N.

Tags are accumulated per batch and applied just before the
post_process_findings_batch task is dispatched, so deduplication and
rules tasks see the tags already written to the DB.

Both default_importer and default_reimporter use the same approach.
For the reimporter, finding_post_processing accepts an optional
tag_accumulator list; when supplied, tags are accumulated rather than
applied inline (backward-compatible for any direct callers).