Remove cudf._lib.nvtext in favor of inlining pylibcudf

[mroeschke]

Description

Contributes to #17317

Checklist

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

[vyasr]

One general thought for future improvements, but overall looks good.

[vyasr]

I wonder if it's worth capturing the general pattern of:

return type(self).from_pylibcudf(
    getattr(plc, func)(
        self.to_pylibcudf(mode="read"),
        *args,
        **kwargs,
    )
)

into a helper function is worthwhile. In addition to reducing boilerplate, it could also be useful to inject standard bits error handling or other pylibcudf wrapping that we always want in place. It may be harder to placate mypy with that kind of code though.

[vyasr]

In the long run I'd envision this function also being able to handle things like _with_type_metadata, for example.

[mroeschke]

Yeah the to/from_pylibcudf pattern is pretty common. I think it may be worth capturing in a helper function, but it may also depend on how much I can make the current cudf Column just be a pylibcudf Column directly

[mroeschke]

/merge