Bug 2034671 - Ingest suite-level replicates from performance data.

[gmierz]

This patch adds suite-level replicates ingestion for performance data. A unit test is also added for it.

[florinbilt]

Hi, thanks for the patch! The change looks functionally correct. I have a few suggestions that I think would make it a bit cleaner.

1. The new block in treeherder/etl/perf.py (lines 230-245) is almost identical to the existing subtest replicates handling (lines 318-333). Would you consider extracting a small helper to keep both call sites in sync going forward?
2. if replicates and len(replicates) > 0 — the len(...) > 0 part is redundant since an empty list is already falsy. Just if replicates: would be enough.
3. Would logger.warning (or logger.error) be more appropriate here?
4. The new test covers the happy path nicely. Could we also add a quick assertion for the case where suite["replicates"] = []?
5. I'm not entirely sure what the correct way to do this is, but in the subtest flow, replicate ingestion happens before should_mark_as_multi_commit, while in the new suite-level flow it happens after MultiCommitDatum.objects.create(...).

Overall, the patch looks functional and these are mostly stylistic suggestions, none of them are blockers. Thank you! r+