Improve Polars docs

[bdice]

Description

This PR improves the docs by reducing the size of the Polars heading (too many words) and tightening up the writing of the docs page.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[bdice]

We can target branch-24.08 or branch-24.10 with this PR (it would be merged after #16813). Reviewers, please weigh in on that decision.

[galipremsagar]

I'm okay with merging into 24.08

[quasiben]

Looks great!

[wence-]

Thanks @bdice

[wence-]

The docs builds are failing because the artifact download is attempting to grab 3.12 artifacts, but those don't exist:

https://github.com/rapidsai/cudf/actions/runs/10950652331/job/30437458677?pr=16820#step:9:139
https://github.com/rapidsai/cudf/actions/runs/10950652331/job/30437458677?pr=16820#step:9:401

I don't know what part of the workflow is controlling the environment variable RAPIDS_PY_VERSION

[wence-]

Oh, is it due to this? rapidsai/ci-imgs@0b91078

[bdice]

Yes, it's due to latest tags changing. I'll see what to do here.

[bdice]

I pinned the CI images to a Python 3.11 version. We'll need to undo that change when this is forward-merged.

[bdice]

The docs render fine.

However, something strange happened with the ARM CUDA 12 CI jobs and they spat out 114 megabytes of logs, seemingly to do with FNV hash functions in Numba. I'm unsure why this would erupt into so much output... I have no explanation, so I will rerun CI and hope for the best.

edit: It looks like the same thing is happening again. This can be investigated next week.