[FEA] Make pylibcudf test suite adhere to its testing philosophy

**Is your feature request related to a problem? Please describe.**
Several of the existing tests in pylibcudf do not follow the testing philosophy outlined in the developer documentation:
https://docs.rapids.ai/api/cudf/stable/pylibcudf/developer_docs/#testing

May include tests that:
- Duplicate coverage already provided by libcudf
- Hardcode input/output values rather than validating against an equivalent pyarrow or pure Python implementation.
- Lack proper fixture-based parametrization using shared dtype fixtures defined in `conftest.py`
- Inefficiently allocate GPU data within the test function rather than using module-scoped fixtures.

https://docs.rapids.ai/api/cudf/stable/pylibcudf/developer_docs/#testing

**Describe the solution you'd like**
Check over the existing pylibcudf test suite and bring it in line with the documented testing philosophy.

**Describe alternatives you've considered**
We could leave the test suite as-is, but it makes it harder for new contributors to follow the intended testing patterns and for reviewers to enforce consistency.