Adding new class for wrapping typical ADVI workflow

[mcnaughtonadm]

This PR will address #15 .

To be able to utilize a template JSON file with the emll code. It is necessary to break the components needed to run the ADVI into their own methods. Here we break the scripts present in the run_hackett_inference.py into a separate class called BMCA (can be renamed to whatever, this is just a placeholder) that loads the cobra model, along with the paths to the various data sources.

[mcnaughtonadm]

The line:

emll/src/emll/bmca.py

Line 72 in cfe80dc

self.model = cobra.io.load_yaml_model(model_path)

should be rewritten to accept any type of cobra model input.

[djinnome]

This is a great start! Let's generalize this a bit so that we can swap out the variational inference method.

Let's also create similar classes for upstream preprocessing and downstream post-processing