Speed up uuid.UUID string parsing

[esadomer]

uuid.UUID.__init__() normalizes every string input by removing urn:, uuid:, braces, and hyphens, even for documented inputs where only one of those operations is needed.

The common documented forms are:

• 12345678123456781234567812345678
• 12345678-1234-5678-1234-567812345678
• {12345678-1234-5678-1234-567812345678}
• urn:uuid:12345678-1234-5678-1234-567812345678

A small refactor can handle the documented urn:uuid: prefix directly with str.removeprefix() and keep the existing broader normalization as a fallback for legacy accepted inputs.

On a local Windows debug build, a same-process benchmark comparing origin/main with the patched Lib/uuid.py showed improvements for the documented forms:

plain_hex:  11584.5 ns -> 11366.8 ns (1.9%)
canonical:  12424.6 ns -> 12253.1 ns (1.4%)
braced:     13293.5 ns -> 12981.1 ns (2.3%)
urn:        13800.0 ns -> 12912.8 ns (6.4%)

I also verified test_uuid passes on the same debug build.

Linked PRs

• gh-150226: Speed up uuid.UUID string parsing #150227

[picnixz]

1. it does not make sense to use a debug build for benchmarks
2. 1% speedup is not the best IMO. Even 6% is not that much IMO
3. I am in the process of rewriting uuid in C so this could help more there.

[picnixz]

I also find the new code less readable as we are still falling back to the original parsing if something went wrong which does not make sense.