Binned NSBI model - parameterized histogram templates

[JaySandesara]

More advanced, long-term task, focusing on statistical inference. Implement a new hybrid SBI-traditional parameter fitting strategy, where the students will develop an alternate model class that a) automatically builds binned templates of the parameterized likelihood ratios on-the-fly during fitting, where the binning algorithm can be developed by students and b) performs the fit using the binned templates from a. The fitting algorithm in b. It needs to be computationally and memory efficient in order to be practical.