Expand format conversion options

[aimalz]

Currently the quantile and histogram approximations can only be made from a parametric truth. It would be easy to support creation of quantile and histogram approximations from samples.

Also, there is currently no support for sampling the gridded approximation (an oversight on my part), which will be fixed as part of resolving this issue.

These features will be necessary for #64 to be useful, unless #51 happens first.

[aimalz]

For the sake of #79, I'm reopening this issue. Copying from #67, the following conversions must be implemented before closing this issue:

• [ ] gridded --> quantiles
• [ ] samples --> quantiles
• [ ] histogram --> quantiles
• [ ] gridded --> histogram
• [ ] samples --> histogram
• [ ] quantiles --> histogram
• [ ] quantiles --> mixture model
• [ ] histogram --> mixture model

In extending this functionality, it would not be hard to also address some other items from #79:

• [ ] implement consistent syntax for conversions by extending the using keyword to some kind of general to/from framework
• [ ] eliminate qp.PDF.truth in favor of using mix_mod and qp.PDF.first
• [ ] rename histogram parametrization to something like stepfunc to reduce confusion
• [ ] remove checking if parametrization exists/calculating with default parameters in favor of exceptions

[aimalz]

Just leaving a comment to bump this up the list of recently updated issues, as there has been an explicit request for the samples --> quantiles conversion to be implemented (and it's one of the most straightforward of the ones on the list). Thanks @sschmidt23 for bringing it up! I will address this before version 1.0 is released, so after submitting the paper to AJ but before it gets through the refereeing process.