Vivarium

- Fix bug in copying a `RandomnessStream` with a new key
- Add documentation of randomness in vivarium
- Add validation to `LookupTable`, `InterpolatedTable`, `Interpolation`, and
`Order0Interp`
- Fix bug writing invalid artifact keys
- Fix `EntityKey` `eq` and `ne` functions
- Remove dependency on `graphviz`
- Move `get_seed` from `RandomnessStream` to `RandomnessInterface`
- Remove `random_seed` from output index and add `random_seed` and
`input_draw` to output columns
- Raise a `PopulationError` when trying to access non-existent columns in a
`PopulationView`
- Fix validation issues in Travis config
- Fix typing issues in `ComponentManager` and `Event`

- Bugfix in population type conversion.

- Bugfix in artifact configuration management.
- Bugfix in population query.

- Be less restrictive about when get_value can be called.

- Clean up event emission.
- Make events immutable.
- Stronger validation around model specification file.
- Move the data artifact from vivarium public health to vivarium.
- Update the ConfigTree str and repr to be more legible.
- Be consistent about preferring pathlib over os.path.
- Add some ConfigTree specific errors.
- Refactor ConfigTree and ConfigNode to remove unused functionality and
make the interface more consistent.
- Extensively update documentation for configuration system.
- Restructure component initialization so that **all** simulation components
are created at simulation initialization time. Previous behavior had
sub-components created at setup time.
- Introduce lifecycle management system to enforce events proceed in the
correct order and ensure framework tools are not misused.
- Remove results writer.
- Overhaul simulation creation to be significantly less complex.
- Update privacy levels for simulation context managers.
- Update context creation and usage tutorials.
- Ditch the 'omit_missing_columns' argument for PopulationView.get. Subviews
should be used instead.
- Consistent naming for rates in data, pipelines, and configuration.
- Introduce resource management system for users to properly specify
component dependencies for population initialization.
- Switch age_group_start and age_group_end to age_start and age_end, making
the naming scheme for binned data consistent.
- Use loguru for logging.
- Fix a bug in transition probability computation.
- Raise error when component attempts to update columns they don't own instead
of silently ignoring them.
- Use consistent data bin naming to make using lookup tables less verbose.
- Rename value system joint_value_postprocessor to union_postprocessor.
- Docs and concept note for values system.
- Be consistent about manager naming on builder interfaces.
- Updated concept docs for entry points.
- Lookup table docs and concept note.
- Bugfix in randomness to handle datetime conversion on Windows.
- Constrain components to only have a single population initializer.

- Bugfix to prevent component list from not including setup components during setup phase.
- Bugfix to dot diagram of state machine.