Bump minimum Python version to 3.12, add ability to calculate periodic table period

[wtbarnes]

This bumps the minimum supported Python version to 3.12 in accordance with SPEC0 and adds testing for Python 3.14.

To support Python 3.14, this PR also gets rid of the dependency on the mendeleev package which was previously being used to compute the period of a given element. The reasoning for getting rid of this dependency was twofold: 1. mendeleev was not compatible with Python 3.14 and 2. the function used for looking up the period was considerably slower than this simple implementation. In the end, it didn't make sense to pick up this dependency when only a very small part of the package was being used.

This PR makes two API changes: a function in fiasco.util.periodic_table_period which computes the period for a given element and a new property period on Ion that provides the period of the element of that ion.

[wtbarnes]

It seems as though something has changed from Python 3.13 to 3.14 that causes the contribution function comparison tests to fail for a few of the example ions.

Looking a little more closely at Fe IX 171.073 Å, it seems this is due to differences in the level populations calculation. As an example, here is the level population as a function of temperature for that transition at a constant density of $10^{10}$ $\mathrm{cm}^{-3}$ as calculated using Python 3.13

and here is the same calculation done using Python 3.14

I do not yet know why these are different or why these would only yield such drastically different results for only a few transntions.

These failures seem to persist across all supported versions of the database for Fe IX 171.073 and 188.496. In v10 and v11, the Ca XV 200.972 test also fails.

[wtbarnes]

This was a "fun" exercise. It turns out that, for whatever reason, in Python 3.14, applying an in-place operation to a matrix in which I'm also applying an operation to the indices also applies an in-place operation to those indices, e.g. if A is a matrix and i,j are indices,

A[i-1,j-1] += 10

will also modify i and j in place. Why this happens in 3.14 and not previous versions of Python I do not know.

Weirdly, I don't seem to be able to reproduce this in any simple way. It could be due to how I'm returning the indices being used from the database.