fix: Set DatabricksConnectionConfig to only use a shared_connection with U2M OAuth

[davem-bis]

Description

The DatabricksConnectionConfig was set to use a shared connection pool for all authentication types to resolve a race condition with U2M OAuth, however this has the side effect that SQL execution is sequential regardless of the concurrent_tasks config setting.

This change updates:

• The shared_connection variable to be a computed field for all ConnectionConfig classes.
• The shared_connection variable to return True only for U2M OAuth authentication methods for the DatabricksConnectionConfig class.

Partially resolves #5858.

Test Plan

• Unit testing for all 3 situations in Databricks.
• Tested on Python 3.12.3, SQLMesh 0.253.3,:
  • databricks-sql-connector 4.0.5 with Access Token and U2M OAuth.
  • duckdb 1.5.3 with a local database.

Checklist

• I have run make style and fixed any issues
• I have added tests for my changes (if applicable)
• All existing tests pass (make fast-test)
• My commits are signed off (git commit -s) per the DCO

[StuffbyYuki]

@davem-bis Thanks for the PR! Could you look into the merge conflict and failed ci tests?

[StuffbyYuki]

@davem-bis Could you look into failed ci tests?

[davem-bis]

Hi @StuffbyYuki - will do, I'm still working on the branch, didn't mean to push the changes yet!

[davem-bis]

Hi @StuffbyYuki - I've should have resolved the test failure from the CI.

I haven't been able to get all of the tests to pass locally, but I have at least managed to get the number of failed tests to match that on main! FTAOD, the tests I have failing locally all look to be passing in the CI.

Once I've got a dev container that can run all of the tests locally I'll submit another PR with that in.

[StuffbyYuki]

I reviewed the pr. Not a blocker, but maybe align U2M detection with _static_connection_kwargs using not oauth_client_secret instead of oauth_client_id is None? Let me know what you think