Some updates for NegativeBinomial

[WarrenWeckesser]

• Uncomment the check_discrete_distribution tests.
• Expand the documentation--explain the meaning of r and p.
• Note that r can be real.

Closes gh-115.

[boxtown]

I've long forgotten the context to #115 but this seems reasonable and if the tests pass then I think it's fine