replace manual conversion factors with `isqx` and annotate physical quantities with it

[abc8747]

Conversion factors between units are currently scattered throughout the codebase. Followed by aerocore.types, fr24.types.isqx, we can use isqx (a zero dependency units/quantity kinds library) to:

import isqx
import isqx.usc

# 1. safely convert between metric and imperial units
fpm_to_mps = isqx.convert(isqx.usc.FT * isqx.MIN**-1, isqx.M * isqx.S**-1)
assert fpm_to_mps(1300) == 6.604

# 2. also embed metadata into type annotations
from typing import TypeVar, Annotated
import isqx.aerospace

_T = TypeVar("_T")

PressureAltitudeM = Annotated[_T, isqx.aerospace.PRESSURE_ALTITUDE(isqx.M)]

def _(alt: PressureAltitudeM) -> ...:
    ...

The use of Annotated[_T, M] and typing.get_type_hints(..., extra=True) to extract M naturally integrates with #37.

cc @xoolive Testing the static analyser here will be fun, though it is very low priority :)

[abc8747]

Quick note on #19 (comment) towards eventual pyright strict mode:

Most of our core aero functions are typed with float | np.ndarray. One may be tempted to do something like:

T = TypeVar("T", float, np.ndarray)

def f(x: T) -> T:
    ...

but this causes many problems. To avoid overcomplicating things, isqx supports various levels of strictness:

_T = TypeVar("_T")
M = Annotated[_T, isqx.M]

# we should prioritise doing this
def generic(x: M) -> M:
    # pyright resolves the type to the internal `Unknown` type, which is equivalent to leaving the method untyped
    # so it will actually try its best to trace/infer the type depending on the implementation and inputs.
    # not to be confused with `typing.Any`, which destroys the type and should be avoided at all costs!

# otherwise if the return value is definitely owned by numpy:
def numpy_facing(x: M) -> M[np.ndarray]:
    return np.asarray(x)
    ...

# or if it definitely returns a scalar float  
def scalar(x: M) -> M[float]:
    return float(x)

For numpy-owned state (traffic.{lat,lon,cas}, conflict arrays, route arrays...) we should just type as np.ndarray (with isqx if appropriate)

See aerocore/airspeed.py for the intended usage.