Add PostgreSQL as a backend target (MVP)

[nj1973]

This issue is to implement PostgreSQL as a backend target but not spend too much effort tuning the PostgreSQL load from GCS, this is because we suspect we need a GCS to PostgreSQL FDW extension, hence the MVP in the title.

For MVP transport my initial thought was to use a single COPY FROM STDIN command but COPY does not accept Avro or Parquet. Is there an alternative or do we need to stage to CSV? Would CSV limit supported data types?

Perhaps instead of MVP this should just be "phase 1" which is to implement everything except data copy?

Tasks:

• Look at connnectivity, we need to support core PostgreSQL, Cloud SQL and AlloyDB. Test TLS too
• Add code to map canonical columns to PostgreSQL data types (+ tests)
• Consider backend table creation, partitioned table creation is more involved than for other backends
• Implement BackendApi methods for PostgreSQL (+ tests)
• Implement BackendTable methods for PostgreSQL (+ tests)
• Implement Predicate Offload methods
• Transport, maybe?
• Implement methods for final and staged validation

I've probably missed some tasks so don't rely solely on the list above.

[nj1973]

Possible alternate flow is for Spark to stage to an unlogged table in PostgreSQL instead of Cloud Storage and then INSERT/SELECT from there.