Make u.Magnitude hashable

[olebole]

Description

Currently, u.Magnitude is not hashable, different from u.mag:

>>> import astropy.units as u
>>> mags = u.Magnitude([1., 2., 3.])
>>> mags
<Magnitude [1., 2., 3.] mag>
>>> mags.physical
<Quantity [0.39810717, 0.15848932, 0.06309573]>
>>> set(mags)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: unhashable type: 'Magnitude'

It would be useful to have magnitudes hashable, to allow them f.e. as a parameter of a function decorated with @functools.cache.
The workaround of using u.mag here is ugly, since it lacks the physical property:

>>> import astropy.units as u
>>> mags = [1., 2., 3.]*u.mag
>>> mags
<Quantity [1., 2., 3.] mag>
>>> set(mags)
{<Quantity 1. mag>, <Quantity 2. mag>, <Quantity 3. mag>}
>>> mags.physical
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'RegularFunctionUnit' object has no attribute 'physical'

Specifically I have

    @functools.cache
    def foo(self, mag, …):
        return … * mag.physical

[nstarman]

I think this is NumPy's fault. I don't think Quantity is hashable, because ndarray is not... ?

[nstarman]

@mhvk, correct me if I'm wrong.

[olebole]

As in my example, a (non-array) Quantity is hashable. It works with u.Quantity, but not with u.Magnitude.