CCL and mass conversions

[anicola]

This PR adds CCL to the likelihood and also allows for more flexibility in mass conversions. The I/O is also refactored.

[itrharrison]

Thanks for doing this! It looks as though CCL is used in a couple of ways

• for the functions available through ccl.haloes
• for getting various background and a pk_interpolator object

For the second one... is there a reason you can't use those which would be already available from the BoltzmannBase theory code, or CCL calculator theory already implemented? I think at the moment different Boltzmann codes could be used in this likelihood to other likelihoods, which seems undesirable.

Similarly for the first one, it would be nice to have the code which implements this as a Theory available to other likelihoods, rather than only in this one. Could it be added to the CCL calculator theory?