🌸🍒 mdsequence cherries

[baum]

PostgreSQL-related performance optimization.

Those changes were 🌸🍒 cherry-picked from dannyzaken/noobaa-core/tree/danny-mdsequence

• Use PostgreSQL sequence
  • currently mdsequence is using a counter inside a jsonb field, which has very bad performance in parallel writes.
  • changed to use postgres native SEQUENCE.
• Increase PostgreSQL client connections pool size if LOCAL_MD_SERVER
  • for the endpoints, we saw that when a large number of small writes is done in parallel, a lot of queries are waiting for clients. the DB has max connections of 600, so it should be able to handle more connections.
• Remove sort by version_enabled in find_object_null_version
  • For some reason, the generated query is causing Postgres to perform a sort according to version_enabled, even though the index is also on version_enabled.
  • version_enabled is always null in this case, and the sort is redundant. removing it solved the issue.