Int input to atmos.rays/compute_rays

[Goobley]

If an int is passed into atmos.rays instead of a float, e.g. due to the case of ctx.compute_rays(wave, 1), it is not caught by the non-iterable check, as this checks isinstance(muz, float). This cascades onwards causing a number of issues, resulting in failure to create a new Context.
Further if one instead puts a list with a singular int, different but similar issues arise due to types getting carried over somewhere, and being incompatible with the backend memory views (e.g. ctx.compute_rays(wave, [1])).

The best fix here is likely to be using something like np.asarray(muz, dtype=np.float64) instead of the current checks (and then ensuring that it's 1- and not 0-dimensional).

For now this issue can be worked around by only passing floats to the mu argument: e.g. ctx.compute_rays(wave, 1.0).