[ENH] Further Improve HMM Estimator

[miraep8]

Currently HMM Estimator has been implemented (#2855) ! :) However there are some changes/improvements I plan to add. Will add and check things off as I get to them:

• add Baum-Welch logic to _fit so that the parameters which are used in the Viterbi algorithm can be dynamically extracted from the data.
• Expand support to multi-dimensional data (should be a simple extension!)
• Move beyond the assumption of equal time spacing
• Add support for partially observed data
• For a "fitted" HMM add the option to "sample" from the model.
• Write a wrapper for hmmlearn
• write an example use case/notebook of how to use it
• Look at implementing other statistical methods eg CRF

FYI @lmmentel