enhancement: Interpolate_by support float in the by column + extrapolate_flat option

[agossard]

This addresses https://github.com/pola-rs/polars/issues/16794. Floats are now supported in the by expression of interpolate_by, and I also added the ability to extrapolate in the regions before and after the start of the by column, by just extending the lowest/highest value of the expressions (consistent with numpy interp function defaults, I believe).

This is my first PR, so I'm not totally sure what's going on. Running the test suite yields many failures, but most of them seem very unlikely to be related to my changes.

One test that is failing in test_interpolate_by. It is not correctly raising the error that it is looking for here:

s = pl.Series([1, None, 3])
by = pl.Series([1, 2])
with pytest.raises(InvalidOperationError, match=r"\(3\), got 2"):
    s.interpolate_by(by)

But I am having trouble understanding where this code goes or debugging it, tbh. Anyway, hoping someone can help.

[MarcoGorelli]

thanks @agossard !

Could you split this PR up into two please?

• support float in the by column
• add extrapolate_flat option

[agossard]

Sure. Will try to work on it tonight.