Added convergence training stopper

[aymanzay]

Added convergence-stopping capability to learn_model() command. When the --converge flag is given, number of iterations is automatically set to 1000 so the model has an ample number of iterations to converge. The tolerance value is configurable to give users more flexibility with when they want the model to stop training.

---

Newly Added CLI Parameters

• --converge: (boolean) flag parameter that triggers an auto-stopper if log-likelihood difference between every 5 iterations is <= tolerance.
• --tolerance -t: (float) value to compare with the log-likelihood difference. If difference is <= tolerance, model training will be stopped.

[codecov[bot]]

Codecov Report

Merging #68 into simplify-model-summary will decrease coverage by 0.79%. The diff coverage is 22.22%.

@@                    Coverage Diff                     @@
##           simplify-model-summary      #68      +/-   ##
==========================================================
- Coverage                   86.48%   85.69%   -0.80%     
==========================================================
  Files                           8        8              
  Lines                         592      629      +37     
  Branches                      129      142      +13     
==========================================================
+ Hits                          512      539      +27     
- Misses                         46       52       +6     
- Partials                       34       38       +4     

Impacted Files	Coverage Δ
moseq2_model/helpers/wrappers.py	86.66% <0.00%> (-1.67%)	:arrow_down:
moseq2_model/train/util.py	88.59% <0.00%> (-4.07%)	:arrow_down:
moseq2_model/cli.py	94.82% <100.00%> (+0.18%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update abe41e2...0dfd617. Read the comment docs.