[python] Ingestion performance

[johnkerl]

There are opportunities within the ingestor to try/catch create then open, rather than the other way around, at zero functional change but with improved processing times.

Idea:

• There are times we want to open something for write if it exists, or create it if it doesn't
  • Sometimes we simply do that: just to get a thing existing at a given URI, whether by creating or opening for write
  • Other times it's an error if the thing already exists (e.g. ingest without ingest_mode="resume")
• Currently we've been doing a try-open-for-write, and catching a does-not-exist error with trying a create
  • All else being equal this would be as good as doing try-create, catching an already-exists error with trying an open for write
  • However in certain contexts, the if-exists logic in the try-open-for-write case can take unnecessarily long
  • This is particularly relevant for ab nihilo ingests for which the thing not existing is the normal case
  • In either ab nihilo ingests, or append-mode ingests, it turns out to be quite performant to do the try-create first and then fail back to try-open-for-write
  • In summary: we have create and open, and open can be more expensive, so it's always best to try create then open

[sc-44853]