Pronto

Fixed
- Issues with typedef serialization in `FastoboSerializer`.
- `Ontology.create_term` and `Ontology.create_relationship` not raising `ValueError`
when given an identifier already in the knowledge graph.
- Signature of `BaseSerializer.dump` to remove `encoding` argument.
- Missing `__slots__` in `Entity` in non-typechecking runtime.
Changed
- Bumped `fastobo` requirement to `v0.6.0`.

Added
- Extraction of `oboInOwl:consider` annotation in `RdfXMLParser`.
- Extraction of `oboInOwl:savedBy` annotation in `RdfXMLParser`.
- Extraction of `subsetdef` and `synonymtypedef` as annotation properties in
`RdfXMLParser`.
- Support for `doap:Version` instead of `owl:VersionIri` for specification
of ontology data version.
- Proper comparison of `PropertyValue` classes, based on the lexicographic order
of their serialization.
- `Ontology.dump` and `Ontology.dumps` methods to serialize an ontology in
**obo** or **obojson** format.
Fixed
- `Metadata` not storing optional description of ID spaces if any.
- Wrong type hints in `RelationshipData.equivalent_to_chain`.
Changed
- Added type checking to some more property setters.
- Avoid using `networkx` in `Term.subclasses`.
- `fastobo`-derived parsers will not create a new entity if one exists in the
graph of dependencies already.
- Exposed `pronto.warnings` and the complete warnings hierarchy.

Added
- Support for extraction of relationships from OWL/XML files to `OwlXMLParser`.
Fixed
- Type hints of `RelationshipData.synonyms` attribute.

Changed
- Dropped support for Python earlier than `3.6`.
- Brand new data model that follow the OBO 1.4 object model.
- Partial OWL XML parser implementation using the OBO 1.4 semantics.
- New OBO parser implementation based on `fastobo`.
- Imports are properly separated from the top-level ontology.
- `Ontology.__getitem__` can also access entities from imports.
- `Term`, `Relationship`, `Xref`, `SynonymType` compare only based on their ID.
- `Subset`, `Definition` compare only based on their textual value.
Added
- Support for OBO JSON parser based on `fastobo`.
- Provisional `mypy` type hints.
- Type checking for most properties in `__debug__` mode.
- Proper `repr` implementation that should roundtrip most of the time.
- Detection of file format and encoding based on buffer content.
Removed
- OBO and JSON serialization support (for now).
- `Term.rchildren` and `Term.rparents` and stop making direction assumptions on relationships.