Hypothesis tests

[wlandau]

As mentioned by @larslau, we will at least want likelihood ratio tests of non-proportionality, as well as significance tests for individual $\theta$ components.

summary() already returns likelihoods, parameter counts, and data sizes, and pmrm_estimates(parameter = "theta") returns estimates and standard errors of $\theta$. We may just be missing degrees of freedom approximations. Satterthwaite seems like a good place to start. Kenward-Roger may be a lot more work.