What this PR does / why we need it:

Another small follow-up to #6965 / #6992 for the local compute engine's Polars backend.

LocalJoinNode converts context.entity_df to a backend DataFrame and then infers the entity timestamp column from that DataFrame's dtypes:

entity_df = self.backend.from_arrow(pa.Table.from_pandas(context.entity_df))

entity_schema = dict(zip(entity_df.columns, entity_df.dtypes))
entity_ts_col = infer_entity_timestamp_column(entity_schema)

infer_event_timestamp_from_entity_df falls back to "the only datetime column" when there is no event_timestamp column, and it detects datetime columns with pandas' is_datetime64_any_dtype. That check does not recognise polars dtypes, so with backend: polars an entity_df whose timestamp column has another name raises:

EntityTimestampInferenceException: Please provide an entity_df with a column named event_timestamp representing the time of events.

The same entity_df works with the pandas backend.

context.entity_df is a pandas DataFrame at this point (it is passed to pa.Table.from_pandas on the line above), so the schema for the inference is now taken from it. Nothing changes for the pandas backend.

A new parametrized test joins against an entity_df with a ts column on both backends; the polars case uses pytest.importorskip("polars"), so it is skipped where polars isn't installed.

Which issue(s) this PR fixes:

No existing issue; found while comparing the local nodes' output between the two backends.

Checks

• I've made sure the tests are passing.
• My commits are signed off (git commit -s)
• My PR title follows conventional commits format

Testing Strategy

• Unit tests
• Integration tests
• Manual tests
• Testing is not required for this change

Local results (Python 3.12):

ruff check, ruff format --check and mypy on local/nodes.py pass.

Misc

With this in, the join, dedup, filter and aggregation nodes gave the same output on both backends for the inputs I compared (with and without an entity_df, several inputs, TTL and expression filters, the supported aggregation functions).