Update solve_lmde_perturbation work with non-integer labels

[DanPuzzuoli]

What is the expected behavior?

Currently the solve_lmde_perturbation function requires that the labels for the perturbations be non-negative integers. It would be nice to allow more general labels, e.g. strings that can directly name what the perturbation represents.

This could either be done by having a translation of arbitrary labels <-> integers before/after the core of the function is executed, or by modifying the actual internals to work with more general labels. The latter would be more elegant, though it'll likely be much easier to do the former. (I can't recall how much of the internals make use of integer properties.)

[DanPuzzuoli]

I was going to try to do this right now but it's not urgent and may make sense to just wait until after arraylias integration.