feat(python): move F-order data in and out of numpy to polars zero copy

[ritchie46]

We could already read F-order numpy data zero-copy into a DataFrame.

This PR adds a fast path for moving data back to numpy. If all arrays are back to back allocated we create a zero-copy numpy buffer.

This means you can move read-only data in and out polars for free. :)

import polars as pl

rows = int(1e8)
cols = 5
x = np.ones((rows, cols), order='F')
# no-copy
df = pl.DataFrame(x)
# no-copy
df.to_numpy()

# But completely safe. We will never mutate
# underlying data. If you do, we will simply
# trigger a copy when you move data out into numpy

Edit: cannot make this work on windows. @ghuls any idea what windows + numpy does differently?

EDIT: nevermind. Using the numpy C-API.