gh-131196: Improve uuid.__str__() performance a little more

[akx]

This PR follows up on the PR #131197 to further slightly (~2%) improve the performance of UUID.__str__() by avoiding trampolining through the hex and bytes properties.

(Given that stringifying UUIDs is likely a pretty common use case, maybe this would be worth a speedup alternative in _uuidmodule.c?)

Here's what a pytest-benchmark comparison says (suuid being a copy of this uuid.py and suuid_orig the one from 3416e7c).

Name (time in ns)	Min	Max	Mean	StdDev	Median	IQR	Outliers	OPS (Mops/s)	Rounds	Iterations
test_suuid[suuid]	370.0099 (1.0)	442.6828 (1.05)	401.2448 (1.0)	20.4759 (2.85)	398.8881 (1.0)	20.2299 (1.88)	3;1	2.4922 (1.0)	10	1,355,472
test_suuid[suuid_orig]	402.7837 (1.09)	422.9409 (1.0)	410.2969 (1.02)	7.1930 (1.0)	408.5021 (1.02)	10.7775 (1.0)	4;0	2.4373 (0.98)	10	1,245,293

Test code

```python import suuid_orig import suuid import pytest

@pytest.mark.parametrize("module", [suuid, suuid_orig])
def test_suuid(benchmark, module):
u = module.uuid4()
def b():
assert str(u)
benchmark(b)

</details>


<!-- gh-issue-number: gh-131196 -->
* Issue: gh-131196
<!-- /gh-issue-number -->

[yihong0618]

FYI: seems there's already one Pull request for uuid
#139123

[akx]

@yihong0618 Ah, neat! That'd cover my other UUID PR #140666 among other things :)

[github-actions]

This PR is stale because it has been open for 30 days with no activity.