Biocantor

Fixed
- Feature interval identifier regex should exactly match qualifier keys

Added
- Unified API for identifiers on all interval objects with new property methods `.id` and .`name`.

Fixed
- `Biotype` enum improperly mapped `protein_coding` and `protein-coding` to different values. Added `mRNA` as another synonym for this type.
- GFF3, BED and GenBank export from Interval objects now raise an exception when the sequence name field is null.

Added
- Parse `FeatureInterval` and `FeatureIntervalCollection` from GFF3 or GenBank, and write back as well.

Changed
- `FeatureInterval` now has multiple types, stored as sets. `FeatureIntervalCollection` stores the union of these types, in addition to optionally having its own type.

Fixed
- `CompoundInterval.relative_interval_to_parent_location()` in the case of overlapping blocks. Had previously been double counting overlap region.
- `CompoundInterval.gap_list()` in the case of overlapping blocks. Had been raising an error in that case.

Added
- `CDSInterval.scan_codon_locations()` method. Returns an iterator over codon locations.
- Implement `__hash__()` for `CompoundInterval` and `CDSInterval`
- Implemented data structures for `TranscriptInterval` and `FeatureInterval` that model transcribed and non-transcribed genomic features.
- Implemented data structures for `GeneInterval` and `FeatureIntervalCollection`, that model groups of intervals as genes or generic feature groups.
- Implemented a wrapper data structure `AnnotationCollection` that contains groups of genes and non-transcribed feature groups.
- Implemented the ability to build BioCantor gene models from GFF3 and GenBank files.
- Implemented the ability to export the above data structures as GFF3, GenBank, BED and NCBI TBL formats.
- Implemented Marshmallow dataclasses that allow for serialization and deserialization of the above data structures.
- Copied the bins implementation from gffutils to avoid needing the full dependency set in a minimal install.
- Added a Biotype enumeration that tracks known biotypes.
- Added caching of sequence retrieval to `Interval` objects.

Changed
- Migrated sphinx documentation from `automodapi` to `autoapi`.
- Performance upgrades to interval arithmetic operations.

Removed
- `CDSInterval.intersect()` method. Frame math was incorrect for complex CDSs and was deemed too difficult to implement correctly.

Fixed
- Remove duplicated code that had been created by a merge issue
- Don't do modular arithmetic with CDSFrame.NONE (value -1)

Added
- Initial release