Add probability module

[lukasheinrich]

Description

another step towards statisfactory. this replaces log_prob(data,pars) calls directly to tensorlib with objects that mirror the API of torch.distributions.Distribution

see e.g. test (which should probably be added here)

https://gist.github.com/lukasheinrich/c69d89333d1c5bd6ea6a0a78eacc919b

Checklist Before Requesting Reviewer

• Tests are passing
• "WIP" removed from the title of the pull request
• Selected an Assignee for the PR to be responsible for the log summary

Before Merging

For the PR Assignees:

• Summarize commit messages into a comprehensive review of the PR

* adds pyhf.probability module
* pyhf.probability : stats classes of backends  :: tensorlib : array classes of backends

[coveralls]

Coverage increased (+0.06%) to 93.819% when pulling 359e7ed on add_probability_module into 57131f7 on master.

[matthewfeickert]

It would be nice to start getting docstrings on both the classes and their methods.

[matthewfeickert]

I removed the docstrings I added and migrated them to PR #558

[matthewfeickert]

We also need a summary commit message for the PR

[matthewfeickert]

We should have some explicit tests of the probability API. I know that this will need to change a bit as the other PRs happen, but I think that it is important to make sure that we are explicitly testing each PR.

[lukasheinrich]

agreed. tests added

[matthewfeickert]

@lukasheinrich Thanks for your patience and for adding tests for probability! LGreatTM!

The only thing left that I will ask is that you

Summarize commit messages into a comprehensive review of the PR

but beyond that I'm happy so I'll approve and leave this to you and @kratsg.

[kratsg]

I would just like some clarification about why we do log_prob for the name. This doesn't seem obvious to me.