On older versions of Python, skip benchmarks that use features introduced in newer Python versions

[AlexWaygood]

All our benchmarks have a requires-python field in their pyproject.toml files, e.g.:

pyperformance/pyperformance/data-files/benchmarks/bm_2to3/pyproject.toml

Line 3 in 974e29c

requires-python = ">=3.8"

The requires-python field is added to the metadata of each benchmark as a python field here:

pyperformance/pyperformance/_benchmark_metadata.py

Line 22 in 974e29c

'requires-python': 'python',

pyperformance/pyperformance/_benchmark_metadata.py

Lines 193 to 202 in 974e29c

	for field, target in PEP_621_FIELDS.items():
	if target is None:
	target = field
	if field == 'url':
	repo = project.get('urls', {}).get('repository')
	raise NotImplementedError
	elif not resolved.get(target):
	value = project.get(field)
	if value is not None:
	resolved[target] = value

We can use that metadata to create a new python property on Benchmark objects, which returns a packaging.specifiers.SpecifierSet instance. This property can then be easily used to filter out benchmarks that require a higher version of Python than the version of Python pyperformance is running on.

Fixes #281. Unblocks #280 and #268.

I haven't added a test for this -- I was unsure if it was necessary, and if it was necessary, where that test should go. I'm happy to add one if that would be helpful and there's an obvious place it could go, however!

I manually tested adding a benchmark that used new-in-Python-3.8 features in 1734919, and the CI passed fine (adding the same benchmark to main currently causes the CI to fail — see #280). Passing CI run on my GitHub fork: https://github.com/AlexWaygood/pyperformance/actions/runs/4813903519

[brandtbucher]

I'm impressed how simple this is! Great job.

One opportunity to remove some branching:

[ericsnowcurrently]

This is a great improvement. Thanks, @AlexWaygood!