GH-112727: Speed up pathlib.Path.absolute()

[barneygale]

Use _from_parsed_parts() to create a pre-joined/pre-parsed path, rather than passing multiple arguments to with_segments()

• Issue: Optimize pathlib.Path.absolute() #112727

[barneygale]

1.5x speedup for paths with tail segments:

$ ./python -m timeit -s "from pathlib import Path" "str(Path().absolute())"
100000 loops, best of 5: 3.82 usec per loop  # before
 50000 loops, best of 5: 3.9  usec per loop  # after

$ ./python -m timeit -s "from pathlib import Path" "str(Path('foo').absolute())"
50000 loops, best of 5: 9.79 usec per loop  # before
50000 loops, best of 5: 6.2  usec per loop  # after

$ ./python -m timeit -s "from pathlib import Path" "str(Path('foo', 'bar').absolute())"
20000 loops, best of 5: 11.5  usec per loop  # before
50000 loops, best of 5:  7.54 usec per loop  # after

It would be interesting to also test the performance of str(WindowsPath('c:foo').absolute()) and str(WindowsPath('/foo').absolute()), but I don't have a Windows box to hand.

[AlexWaygood]

Big speedups on Windows!

On main:

>python -m timeit -s "from pathlib import WindowsPath" "str(WindowsPath('c:foo').absolute())"
Running PGUpdate|x64 interpreter...
50000 loops, best of 5: 9.92 usec per loop

>python -m timeit -s "from pathlib import WindowsPath" "str(WindowsPath('/foo').absolute())"
Running PGUpdate|x64 interpreter...
50000 loops, best of 5: 8.32 usec per loop

With PR:

>python -m timeit -s "from pathlib import WindowsPath" "str(WindowsPath('c:foo').absolute())"
Running PGUpdate|x64 interpreter...
50000 loops, best of 5: 6.01 usec per loop

>python -m timeit -s "from pathlib import WindowsPath" "str(WindowsPath('/foo').absolute())"
Running PGUpdate|x64 interpreter...
50000 loops, best of 5: 4.97 usec per loop

(Timings done using a reasonably fresh PGO-optimised build)

[AlexWaygood]

I can also repro all your reported speedups from #112728 (comment)!

[AlexWaygood]

LGTM

[barneygale]

Thanks tons!