Add a FloatInt (float | int) type alias - #189

PEP 484's numeric tower makes `int` assignable to any `float`-annotated
parameter, attribute or variable, even under `mypy --strict`, while at
runtime `isinstance(1, float)` is `False`. A plain `float` annotation is
therefore a lie: it silently admits values that fall through an
apparently exhaustive `match … case float():` into `assert_never()`, and
that lack `float`-only methods like `hex()`.

There is no clean fix in Python, and the alternatives were all measured
or analyzed and rejected: coercing at ingress costs ~2.3x on the
construction of hot-path types, structural `Protocol` tricks don't close
the widened variable and `Sequence` covariance holes, and widening
`match` arms one by one leaves the annotation lying. So stop lying
instead and spell out what PEP 484 actually admits, at zero runtime
cost.

The docstring example is mirrored as a regular test because the example
linter only extracts module, class and function docstrings, so an
example documenting a module attribute is never checked.

Signed-off-by: Leandro Lucarella <luca-frequenz@llucax.com>