ExternalDistributions

This introduces several new distributions that are constructed based on externally specified samples. Users will need to specify a path to the samples along with mappings between variate names and the columns in the samples (e.g., mass1='m1_source'). The two new class, then, either

• sample from the external samples with replacement and model the pdf as a sum of delta-functions at each external sample (ExternalDistribution) or
• construct a weighted Gaussian Kernel Density Estimate based on the samples, sampling from that distribution within _rvs and evaluating it within _pdf (ExternalKDEDistribution).

Users should be able to use this within something like a KKL estimate based on PE from a single event or ingest complex samples generated with external code that (for whatever reason) cannot be conveniently ported to a specific distribution implemented within the library.

Please note! these changes have not been tested!