Better documentation of save and load methods

[wd60622]

What are the requirement of the inference data/ attrs? What methods are required from child classes?

Currently seems like the process is save and load a single InferenceData object. Apart from the usual prior, posterior, etc, this object will have:

• fit_data group which will contain the data used for fitting the model
• attrs attribute which has all the required information to reconstructure the initial model. This only has the restriction of being JSON serializable.

However, in this process, there seems to be much writing in order to load a model as the attrs need to be parsed and the class needs to be reconstructed. Aside from just loading, the model needs to be built.

[wd60622]

There is a bit of information in the deployment section here:
https://www.pymc-marketing.io/en/latest/guide/benefits/model_deployment.html

Think the documentation for the ModelBuilder itself could be spruced up