Refactor fit() function to separate functions. about espei

[bocklund]

I'm thinking about this approach:

1. pass the input settings (either the whole dict or the top level dicts as above) to an (un)wrapper function that creates objects such as file handles, Database and TinyDB objects
2. Unwrapper calls the underlying generate_parameters and/or multiphase_mcmc_fit functions with the correct objects

This might be the best of both worlds because users could load their YAML or pass a dict to the unwrapper and get what they want, but also easily call the functions with existing objects naturally. (e.g. generate parameters in the notebook, then pass the resulting database directly to the MCMC function, rather than writing to file and then loading from file)

This is sort of how the main function works in the run_espei module, this would just formalize that a bit. The unwrapper function would be a function in run_espei and main would basically handle only the kwargs (check datasets vs. an input file) and then handle loading that file and pass off to the unwrapper.

[bocklund]

Part of this is about having a nice jupyter notebook API