ENH: add possibility to fix parameters in multivariate normal fit

[dschmitz89]

What does this implement/fix?

This is a follow up from #18361: add the possibility to fix mean and covariance during fitting a multivariate normal to data. Such functionality is built in for univariate distributions. The math is pretty straight forward, the tricky parts are useful error messages.

[mdhaber]

Could you point to the reference that the math is as simple as this when one shape is fixed?

[mdhaber]

@dschmitz89 I took another quick look but I wasn't sure that you were done addressing comments. Please ping me when you think this is ready for another look.

[dschmitz89]

Could you point to the reference that the math is as simple as this when one shape is fixed?

@mdhaber : this should be ready now.

About your question: You mean that the covariance matrix is fixed? That should be covered here. $\mu$ does not depend on $\Sigma$ but $\Sigma$ on $\mu$.

[mdhaber]

Very close. Just a few suggestions.

[dschmitz89]

Thanks @mdhaber , I was about to fix the last errors now but you already went ahead, much appreciated.