feat: prql namespace for polars.DataFrame and polars.LazyFrame

[eitsupi]

Close #371

import polars as pl
from datetime import date
from pyprql import polars_namespace

df = pl.DataFrame({
    "a": [1, 2, 3],
    "b": ["zz", "yy", "xx"],
    "c": [date(1999, 12, 31), date(2010, 10, 10), date(2077, 8, 8)],
})

df.prql.query("select {a, b, c_year = s'EXTRACT(year FROM c)'} | filter a > 1")
# shape: (2, 3)
# ┌─────┬─────┬────────┐
# │ a   ┆ b   ┆ c_year │
# │ --- ┆ --- ┆ ---    │
# │ i64 ┆ str ┆ i32    │
# ╞═════╪═════╪════════╡
# │ 2   ┆ yy  ┆ 2010   │
# │ 3   ┆ xx  ┆ 2077   │
# └─────┴─────┴────────┘

[max-sixty]

Looks cool!

No objection to starting this as an optional dependency. We can then promote it if it turns out to be used lots. (I had a bit more objection in #374 to making everything optional).

Another option is to detect whether polars is installed in pyprql.__init__ and add the namespace without having to import pyprql.polars_namespace — depends on the polars norms.

[eitsupi]

Another option is to detect whether polars is installed in pyprql.__init__ and add the namespace without having to import pyprql.polars_namespace — depends on the polars norms.

That seems like a good idea, but I don't know if it is good practice to do that here.
For now, I would like to merge and release it as is.

[max-sixty]

For now, I would like to merge and release it as is.

For sure

That seems like a good idea, but I don't know if it is good practice to do that here.

It would be something like this in the __init__.py:

try:
    import polars
except ImportError:
    pass
else:
    # Add to namespace