Amsterdam-schema-tools

* Support loading and validating publishers from the schema-server.
* Make schematools aware of the metaschema major versions it can work with.
* Support for attempting validation against multiple metaschemas.

SUPPORTED METASCHEMAS: 1 2

* Several minor fixes to tests.
* Removal of unused DatasetSchema.identifier property
* Add `neuronId` is mapping needed for through table identifiers

* Mocked schemas now use properly camel-cased field names.
* Relations can be primary keys.
* The command `schema batch-validate` now works on table files as well as
`dataset.json` files.

* Fix importing schema files by using a relative path.
* Fix `related_dataset_schema_ids` to also detect changes in nested objects.
* Fix `DatasetTableSchema.get_fields()` to return cached instances too.
* Fix `verbose_name` of `GeometryField` in Django ORM, which reused globally defined data.
* Fix performance of iterating over subfields, no longer needs to load related tables.
* Added `DatasetFieldSchema.is_nested_object` property.
* Normalized exceptions for missing datasets/tables/fields:
* The `DatasetNotFound` exception extends from `SchemaObjectNotFound`.
* Added `DatasetTableNotFound` and `DatasetFieldNotFound`.
* There is no need for `except (DatasetNotFound, SchemaObjectNotFound)` code, it can all be `except SchemaObjectNotFound:`.
* Cleanup Django model field creation logic.
* Cleanup SQLAlchemy column creation logic.
* The schema validator now rejects tables with both an 'id' field and a composite primary key.

* Fix `limit_tables_to` issue with crash in index creation for skipped tables.
* Fix `limit_tables_to` issue for M2M relations, now reports the table is not available.
* Fix SRID value for SQLAlchemy geometry columns (were always RD/NEW).
* Fix CKAN upload to skip datasets that are marked as "not available".
* Improved 3D coordinate system detection, and added more common SRID values.
* Improved naming of geometry column index to be consistent with other generated indices.

* Fix `BaseImporter.generate_db_objects()` to handle properly snake-cased table identifiers values for table creation.
* Improve the underlying `tables_factory()` logic to support snake-cased table identifiers for all remaining parameters.