Gate the StrEnum fallback on sys.version_info instead of try/except

[5point1]

Changes included in this PR

Compat chore: replace the try/except ImportError StrEnum fallback with a sys.version_info >= (3, 11) gate in the three enums.py modules (v16, v201, v21).

Current behavior

Static type checkers can't tell which branch of a try/except ImportError wins, so the StrEnum base — and with it the type of every enum member — is ambiguous to them. Checking a downstream codebase with ty reports invalid-argument-type every time an enum member is passed to a parameter annotated with its own enum type (e.g. RegistrationStatus.accepted against a status: RegistrationStatus field).

Minimal repro (ty 0.0.61)

try:
    from enum import StrEnum
except ImportError:
    from enum import Enum

    class StrEnum(str, Enum):
        pass


class Color(StrEnum):
    red = "RED"


def f(c: Color) -> None: ...


f(Color.red)  # error[invalid-argument-type]

The same snippet with if sys.version_info >= (3, 11): passes cleanly.

New behavior

None at runtime: enum.StrEnum is still used on 3.11+, and the fallback still covers 3.8–3.10. sys.version_info gates are the idiom type checkers special-case and prune (it's what typeshed uses throughout), so enum members type correctly again. On a downstream CSMS project, ty went from 43 to 24 diagnostics with this patch — all 19 removed were these false positives.

Impact

No breaking change; the import logic is identical at runtime. The # pragma: no cover moves onto the else: clause (ocpp/v16/enums.py still reports 100% coverage locally), and the CI matrix already exercises both branches (3.8–3.10 take the fallback, 3.11–3.13 the stdlib import).

Checklist

1. Does your submission pass the existing tests? — 182 passed, via poetry install + the make tests commands
2. Are there new tests that cover these additions/changes? — no behavior change to cover; both branches are already exercised by the 3.8–3.13 CI matrix
3. Have you linted your code locally before submission? — black, isort and flake8 all clean

[a-alak]

Thought: since < 3.11 stops being supported in October, should this package maybe follow that?
Or are a lot of users expected to be on 3.8-3.10?