Add warnings on CUDA that compile or runtime performance may be affected

[szymonlopaciuk]

Description

Current behaviour of ContextCupy with respect to automatic disabling of expensive optimisations is a bit opaque, and can be transparently enabled when not desired, but also not enabled when expected.

This PR:

• adds an explanation about the auto-disabling of expensive optimisations when supported, and introduces the flag XO_CUDA_NO_FAST_COMPILE and the module-level switch xo.context_cupy.no_fast_compile, used to control this behaviour;
• when XO_CUDA_NO_FAST_COMPILE is not given, but the current NVRTC version does not support disabling the optimisations, emits a warning to the user that they may need to update CUDA/NVRTC if compile performance is a problem.

Additionally:

• this PR changes the CI lint workflow to use pre-commit so that the checks on CI and local match: currently this is not the case, which causes confusion.

Checklist

Mandatory:

• I have added tests to cover my changes
• All the tests are passing, including my new ones
• I described my changes in this PR description

Optional:

• The code I wrote follows good style practices (see PEP 8 and PEP 20).
• I have updated the docs in relation to my changes, if applicable
• I have tested also GPU contexts