GH-101362: Omit path anchor from pathlib.PurePath()._parts

[barneygale]

Improve performance of path construction by skipping the addition of the path anchor (drive + root) to the internal _parts list. Rename this attribute to _tail for clarity.

Also:

• Simplify the implementations of name, parent, and parents a little.
• Optimize _make_child_relpath() by building the new string path with naive string concatination rather than _format_parsed_parts().
• Optimize __hash__() and __eq__() by referring to the unsplit case-normalized path.
• Optimize __lt__() etc by referring to the naively split case-normalized path (i.e. split on sep without consideration of anchor)

The public parts tuple is unaffected.

$ ./python -m timeit -s 'from pathlib import PureWindowsPath as P' 'P("C:/foo/bar.py")'
100000 loops, best of 5: 1.98 usec per loop  # before
200000 loops, best of 5: 1.75 usec per loop  # after

• Issue: Optimize pathlib path construction #101362

[barneygale]

I'm not sure about this change - I don't like how it makes _parts and parts have different contents. Too subtle for future maintainers to pick up on, perhaps?

edit: addressed by renaming _parts to _tail.

[barneygale]

This change is necessary in order to efficiently fix #65238, which is required before we can return unnormalized paths from __fspath__() in #102783.

[barneygale]

Updated timings:

$ ./python -m timeit -s 'from pathlib import PureWindowsPath as P' 'str(P("C:/foo/bar.py"))'
100000 loops, best of 5: 3.92 usec per loop  # before
100000 loops, best of 5: 3.57 usec per loop  # after

Something like a 10% speedup.