Sciunit

Enhancements:
- Added workshop tutorial notebook
- import path included in _class field
- Added RandomScore and more ZScore error checking
- Added only_lower_triangle option to TestM2M
- Added key-based lookup to TestSuite to find tests with a given name
- Added log_norm_score
- Added optimizer kwargs to TestSuite
- Added reasons to incomplete scores

Bug fixes:
- Fixed recursion bug for missing attributes
- Bug fixes for TestSuite
- Fix PipSession import for pip version 20

Other changes
- Changed to global installation in tutorial
- Updated Dockerfile
- Minor documentation improvements
- Removed deprecated pandas Panel class
- Added Python 3.7 and 3.8 to Travis CI integration tests
- Added contributor guidelines, code of conduct, and executive committee roster

Features added:
- Added `ProtocolToFeaturesTest`, a good base class for many `get_prediction` implementations.
- Added pre-execution validation option for test observations
- Added cerberus validation for test parameters
- Added platform (i.e. OS) information to utils module
- Search for model attributes will also search on backends
- Added `log` function into main namespace
- Added `failed_capability_checks` method to models.
- De-duplicated parameters validator code and added `units_map`
- Added schema format and classmethod to get observation schema names
- Simpler "instance-less" capability checking
- Added `condition_model` method
- Added validation of iterable and multi-unit observations
- Added dependent repos badge

Bug-fixes and compatibility changes:
- Fixes for downstream applications that use Python 2
- Added __future__ imports for Python 2
- Updated model capability unit test
- Added missing kwarg in `Test.check`
- Added `require_extra` back
- Added optional `recorded_variables` attribute
- Added Python 2 and 3 compatible temp directory utility function
- Added better support for default test parameters
- Added warning traceback function
- Added dict_combine function for Python 2 compatibility
- failed_extra_capabilities now checks if model claims the capability