Make sure that the functions we use in a likelihood evaluation are static classes

[andreufont]

Copying from the Google doc discusion:

Related to the discussion above about cosmo vs fid_cosmo. The way I understand this function get_chi2 is that it is a static function, like most functions in the likelihood and theory classes. This means that calling that function should not change the state of any of the internal objects within the likelihood class (not the data, not theory, not the emulator).

So if we ask for get_chi2 and provide as a likelihood parameter a cosmological parameter, this will not modify self.theory.fid_cosmo, it will create a new cosmo object with the right settings that will only leave during the call to get_chi2, and then disappear.

[andreufont]

Aha! It looks like we have gone through this before... Let's see if this time we can finish the discussions, and reach a conclusion that we can settle on and document properly.

By the way, the name of the issue is confusing. It should say "Make sure that the functions we use in a likelihood evaluation are constant functions" (so that they don't modify the current state of the likelihood or theory objects)

[andreufont]

This is now addressed in the new_theory branch, about to be merged.