Minor fixes to modelbuilder class

[butterman0]

Made the following fixes:

1. _validate_data method is called for predict_posterior method, but not predict method. Should be called in both as the only difference between the two should be whether mean or full posterior samples are returned.

2. self.is_fitted_ = False set in class init, but not set to True when model is fitted with fit. Decided to keep the variable as it is useful, therefore set to True after 1) fit called 2) load with fitted data.

The alternative is to remove self.is_fitted_ altogether, seeing as the other functions tend to check for fitting via the existence of inference data (model.idata) instead. Alternatively, could refactor these checks to instead use is_fitted_.

[zaxtax]

Looks decent! Make the lint errors go away and the unit tests in model builder pass, and I think we can merge this

[butterman0]

pre-commit.ci autofix

[butterman0]

Thanks @zaxtax, got autofix to do it! Let me know if anything else.

[butterman0]

Hi @zaxtax, I rebased as there seems to be some problems with pytest. Anything else I need to do here?

[butterman0]

Hi @zaxtax, I removed the commit to add the call to _validate_data as I needed it in #426 to fix testing errors. There would have been some hacking going on otherwise. Not sure if this is good practice, but there won't be any conflicts now!