fix(rust, python): Correctly convert data frames to NumPy for C index order

[borchero]

Motivation

#16288 introduced a regression that caused to_numpy(order="c") calls to silently yield erroneous NumPy arrays whenever the data frame has more than one chunk. To reproduce with the latest release or polars main:

import numpy as np
import polars as pl
from polars.testing import assert_frame_equal

rng = np.random.default_rng()
df = pl.DataFrame({
    f"col_{i}": rng.normal(size=100_000)
    for i in range(1000)
})
df_chunked = pl.concat([df.slice(i * 10_000, 10_000) for i in range(10)])
assert_frame_equal(
    df_chunked,
    pl.from_numpy(df_chunked.to_numpy(order="c"), schema=df_chunked.schema)
)

Currently, this results in an assertion failure. This PR fixes the issue.

---

I'm not sure if this is easily feasible but it would be great if this change could be released prior to a v1 release as it is quite a nasty bug 👀

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.01%. Comparing base (1a47c46) to head (768931e).
Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #17000      +/-   ##
==========================================
- Coverage   81.02%   81.01%   -0.02%     
==========================================
  Files        1446     1446              
  Lines      190424   190422       -2     
  Branches     2709     2708       -1     
==========================================
- Hits       154291   154270      -21     
- Misses      35636    35655      +19     
  Partials      497      497              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ritchie46]

Thanks for the fix. Can you add the test with a small amount of data. Something like this:

rng = np.random.default_rng()
df = pl.DataFrame({
    f"col_{i}": rng.normal(size=20)
    for i in range(3)
})
df_chunked = pl.concat([df.slice(i * 10, 10) for i in range(3)])
assert_frame_equal(
    df_chunked,
    pl.from_numpy(df_chunked.to_numpy(order="c"), schema=df_chunked.schema)
)

[borchero]

Thanks for adding the test @ritchie46!