17 run sbi inference on toy data

[maho3]

Major framework additions for inference and validation. Notes:

• Built the framework necessary to load summaries+parameters, pass them into the sbi inference pipeline, and generate simple validation metrics
• Major steps are structured around Runner objects and yaml config files, as in ili-summarizer
• Provided an example_inference.py script to demonstrate inference on a toy problem

Issues:

• Things are still very hacky and 'minimum working example'
• Code perhaps relies too much on the sbi package. We may want to wrap the sbi objects at some point with our own generic classes
• Only one validation metric at the moment, which is just plotting the posterior for a single data point
• There is no setup.py file yet, nor a consolidated list of required packages. We should figure out how we want to compile these, and possibly handle conflicting dependencies

[maho3]

First Constraints!

[maho3]

First rank statistics! Sometimes biases appear at the edges of training data. TODO: correct them.