Mal-toolbox

[Add method 'AttackerAttachment.add_entrypoint' and test](https://github.com/mal-lang/mal-toolbox/commit/68e30e6e577d99b7204b1a6ff9899e3fd1b69b84)

[Add method 'AttackerAttachment.remove_entrypoint' and test](https://github.com/mal-lang/mal-toolbox/commit/1be262638446f7135849607a8cf56ceda00bed9f)

Fix default value of AttackGraphNode.extras to be a empty dictionary instead of list.

Add the 'extras' field for AttackGraphNodes, where extra meta data can be placed.

Fix bug with language graph override of attack steps.

Changes:
- Remove the specification submodule of the language module.
- Functionality moved into LanguageGraph
- Update the securiCAD translator to work with the current version of the model format and structure.
- Partially update the Neo4j ingestor module to be able to extract the model from a Neo4j instance
- AttackGraphNode.extras now defaults to dict instead of None

Bug fixes:
- Fix issue when attaching Models to AttackGraphs when loaded from file where assets may have changed name when trying to link and add model field to AttackGraph
- Assets were previously not attached to attack steps before the mappings full_name_to_node was generated, which created issues. This is now fixed and the mappings work as they should.

- Fix typo in language graph when handling `subtype` step expression (Note: the language graph still does not handle subtypes correctly, that is for the next release)
- Add more debugging info and fix potential bug in `association_exists_between_assets` method
- Add `get_asset_defenses` and `get_association_field_names` model helper methods