Support mixed Spectrum types

[JSKenyon]

This is an extension of current functionality to allow for input models to cope with different spectral index specifications. This can come about as the result of merging sky models from various sources.

The current plan of action is to wait for improvements to tigger-lsm such that it supports additional Spectrum classes. QuartiCal will then need to partition sources by Spectrum type. This can become quite complicated as we are also partitioning by tag/cluster and source type. Practically, we just need to sum over each Spectrum type in addition to the Source type (directions are not summed over).

For reference, the summation will likely need to happen here: https://github.com/JSKenyon/QuartiCal/blob/200779fb895da57310004e29658972d8ec3fb135/quartical/data_handling/predict.py#L748-L773

[JSKenyon]

@sjperkins @o-smirnov @rubyvanrooyen @landmanbester @bennahugo Just making sure everyone that everyone who was present is aware of this one.