Support multivariate scipy.stats continuous random vectors

[jorenham]

Extend the scipy.stats joint/multivariate distributions with L-comoment (ratio) methods, using lmo.theoretical.l_comoment_from_pdf and lmo.theoretical.l_coratio_from_pdf.

This requires the joint PDF, and the marginal CDF's.
Unfortunately, the marginals are nowhere to be found in scipy.stats._multivariate.multi_rv_generic or its subtypes.

So the only way to implement this feature, is by figuring out the marginals of each joint-distribution, and manually add the l_co(moment|ratio|stats|loc|scale|rr|skew|kurtosis) methods to their respective scipy.stats._multivariate.{}_(gen|frozen)
types.

• multivariate_normal (norm marginals)
• dirichlet (beta marginals)
• multivariate_t (t marginals)

The 13 other ones are either discrete, matrix-valued, directional, or have no .pdf() method, and therefore out-of-scope.