[Merged Locally] Further features and tests

[ZedThree]

Builds on #72

New features:

• calling type-bound subroutines: call foo%bar()
• procedure variable declarations
• polymorphic variable declarations, including class(*)
• variable declarations using type with parameterised intrinsic types: type(real(kind=real32))
• renaming in use statements: use foo, only : bar => quux

Changes to existing features:

• abstract interface as a separate named node
  • This would be useful for ford in order to quickly pull out abstract interfaces vs explicit interfaces
• restrict set of attributes on derived type definitions to abstract, public/private, bind, extends
  • It would be better if we could specify these can only appear once each, but tree-sitter explicitly doesn't support permutations
• add class is to select type
• fix class default for select type
  • In #72, the class default case was parsed as a repeated node: (default) (default)

Includes tests for all new and changed features

[stadelmanma]

@ZedThree I merged this locally after doing a rebase to resolve some conflicts with different versions of tree-sitter in parser.c. I attempted to push to the branch after that but I didn't have permission. Doesn't look like GitHub is recognizing the local code merge like GitLab usually does.

(if something looks like it went astray with the local merge let me know)

Thank you for the updates! Out of curiosity what is your main use case for tree-sitter is it the doc generation project you linked?

[ZedThree]

Great, thank you!

Yep, I maintain Ford and we've been investigating ways of pulling the parser out and using a standalone parser. This looks very promising! There's a few other tools in the Fortran ecosystem that might also be interested. Personally, I'd also like to use it for syntax highlighting in Emacs, because the built-in regex-based highlighter struggles with very modern features like module procedure in submodules.