feat(importers): apply import-time tags per batch before post-processing, do not tag old findings

[valentijnscholten]

Summary

• Import-time tags (apply_tags_to_findings / apply_tags_to_endpoints) were previously applied after all findings were processed and post-processing tasks (deduplication, rules) had already been dispatched — those tasks could not see the tags on the findings.
• Tags are now applied per batch inside process_findings(), immediately after parser tags and location persistence, before dojo_dispatch_task() — the same pattern already used by bulk_apply_parser_tags.
• Closed findings are intentionally excluded: they are absent from the current report and should not receive the import tags.
• Renames the reimporter's ambiguous batch_findings (match-batch slice of unsaved parser objects) to unsaved_findings_batch, freeing batch_findings for the new dedupe-batch accumulator and fixing a latent iterator-mutation bug where appending to the iterated list caused extra loop iterations.
• Consolidates apply_import_tags() into apply_import_tags_for_batch(), removing a redundant Finding.objects.filter() re-query and an unused Iterable import.
• Updates performance test query counts accordingly (each import loses 1 query from the removed post-loop apply_import_tags call).

[mtesauro]

Approved