Interpret-community

SDK:
- hotfix for breaking change in latest version of dependency interpret-core perf_dict api

Visualization:
- Categorical cohort fix for explanation dashboard

SDK:
- String class labels fix with mimic explainer
- Add ipython as a dependency to visualization extra
- Fixed credentialed POST requests for predictions in dashboard (tested in notebook VM), and added support for remote VM with public IP
- Update reference to interpret-core to 0.2.1

Test:
- Fix for job retries in pipelines

SDK:
- improvements to docstrings/error messages around large datasets

Viz:
- added support for passing errors out of the dashboard for hosts to work with

Testing:
- improved notebook tests
- removed deprecated test pipelines
- added pypi release pipeline (trying it out for this release)

SDK
- MLflow `log_model`-based implementations of `log_explanation` and `get_explanation`

Dashboard
- update to categorical handling
- rework internals so hotkeys work better in Jupyter notebooks

Internal
- lots of updates to test pipelines
- README improvements

External:
* MLflow functions now accessible from `interpret_community.mlflow` rather than `interpret_community.mlflow.mlflow`
* Multiclass visualization available in the new dashboard

Internal:
* tracking URI in MLflow testing now configurable
* linting for TS visualization code
* websocket-extensions upgrades for security
* additional `explanation_method` information will be passed to the dashboard for display
* pin keras dependency for testing
* added Janhavi as an author!

Visualization
* chart improvements
* support for regression in new dashboard
* tooltips for charts
* additional bugfixes

SDK
* rework of explanation serialization code in preparation for MLflow integration
* experimental MLflow log function added