Lightcurve-fitting

This release adds the `ShockCooling4` model of Morag, Sapir, & Waxman ([2023](https://doi.org/10.1093/mnras/stad899)).

This release includes a major refactor of the `models` module. Models now require initialization with the `.LC` object to be fit. Updated examples are given in the documentation and the Jupyter notebook. This alleviates the need for model keyword arguments, which are now deprecated. (A warning will be issued if the user tries to supply a `model_kwargs` argument.) This also allows for much easier specification of new models by subclassing the existing models. See `release-history.rst` for other minor changes.

This release allows for fitting SEDs with any arbitrary function, not just a blackbody. It also adds support for JWST filters. See other minor changes in release-history.rst.

This release implements the `CompanionShocking2` model for Type Ia supernovae, which replaces the three multiplicative factors in the `CompanionShocking` model with two time offsets in the U and i SiFTO templates. See other minor changes in release-history.rst.

This release implements a new version of the shock cooling model, `ShockCooling3`, which is identical to `ShockCooling` except that it allows distance and extinction to be free parameters. Therefore it fits the flux instead of the luminosity. To allow for Gaussian priors on these parameters, I had to make the prior specification slightly more complex. See other minor changes in release-history.rst.

This release replaces filter transmission functions with generic filter curves from the Spanish Virtual Observatory wherever possible. It also changes the prior on the blackbody temperature from log-uniform to uniform for the purposes of bolometric light curves. Since the previous version, there was a bug in calculating the maximum/minimum validity times for shock cooling models when using an intrinsic scatter parameter, which is now fixed. See other minor changes in release-history.rst.