[PyArrow] Filter Pushdown of IS IN now uses the dedicated pyarrow.compure.Expression.isin method

[evertlammerts]

Continued from #73

Null handling seems to work fine, afaics. We are using Expression.isin(), which does not have a skipp_nulls kwarg. It uses the same default semantics as pyarrow.compute.is_in, which is skip_nulls = False. This is not exactly the same as duckdb's semantics, because DuckDB will not return NULL values even if NULL is in the IN list:

D select * from VALUES(1), (2), (3), (NULL) as t(myint) where myint in (2, NULL);
┌───────┐
│ myint │
│ int32 │
├───────┤
│   2   │
└───────┘

But this is no big deal. We might let more rows through than is needed, but duckdb will filter them out anyway. Also see the test.

[Tishj]

Looks good, thanks for picking it back up! 👍

[jakkes]

Thank you!