feat: Auto schema evolution revisited

[k-anshul]

Revisiting an old feature that allow automatic schema evolution. Our current approach of ingesting data in a new table for partition overwrite and merge strategy makes it much easier to implement now. Thus it is only supported for partition_overwrite and merge strategy based incremental ingestion(ideally we should just remove append strategy but that is discussion for some other issue/alternatively we can modify append strategy to also ingest data in a temporary table first).
This will avoid complicated pre_exec statements that we need to do that which should also ensure that full rebuild can still be supported.
Three strategies are implemented:

1. Fail : Fail if any column is added or removed. Datatype changes are not compared as of now and we let db do automatic translation or fail. Set as default for backward compatibility.
2. Ignore: Ignores new columns but fails if column is removed.
3. append_new_columns: Appends new columns by automatically altering current table.

Note: The implementation is similar to dbt's schema evolution approach but we avoid adding sync_all_columns strategy which also alter data types. It has several nuances which dbt also misses like datatype detected to a narrower datatype than source data, whether auto conversion to string should be allowed, whether we should let duckdb do lossy conversion like timestamp to date type etc. This can be revisited to only allow wider datatype conversions.

Future improvements:

1. Add sync_all_columns strategy as detailed in note above.
2. Also implement for clickhouse.

Checklist:

• Covered by tests
• Ran it and it works as intended
• Reviewed the diff before requesting a review
• Checked for unhandled edge cases
• Linked the issues it closes
• Checked if the docs need to be updated. If so, create a separate Linear DOCS issue
• Intend to cherry-pick into the release branch
• I'm proud of this work!