Support full reconstruction of HCL from parse tree

[weaversam8]

Hi there, this closes #23. Many folks were asking for the ability to convert the parsed files back into HCL. This PR does a big part of the heavy lifting required to enable that functionality.

Currently the ultimate output of this library is a dictionary transformed from the underlying AST generated by Lark. I did not attempt to reverse the transformation process that this library takes while traversing the AST. (Maybe a later PR.) I did, however, use Lark's experimental "reconstruction" functionality to provide a way to transform the AST back into valid HCL. As such, if developers are willing to manipulate the AST instead of the final output dictionary, reversing the parse back into HCL is now possible with this PR.

A high level overview of the changes in this PR:

• I added new API methods:
  • parse(file) and parses(string) return an AST (Lark.Tree) after parsing an HCL2 file
  • writes(ast) writes an AST back to a valid HCL2 string
  • transform(ast) allows a user to convert the AST to a dictionary using the existing DictTransformer class, like the load and loads methods already do under the hood.
• I modified the Lark grammar to capture whitespace within newline, comment, and = tokens, which doesn't change how those rules are ultimately parsed
• I made a few small changes to DictTransformer in hcl2.transformer to ensure it worked with the tweaks I made to the grammar
• I implemented hcl2.reconstructor, which rebuilds the parser without a cache, implements the Reconstructor, and contains logic to add whitespace to the correct tokens when reassembling the file
  • (I can't load the parser from a cache due to this Lark issue: cache option does not work with Reconstructor lark-parser/lark#1472)
  • The parser is rebuilt the first time one of the new API methods are called, so the performance benefits from caching should remain in place for existing users of this library
• I added new tests in test/unit/test_reconstruct.py to test my new functionality.
  • test_write_terraform ensures that every HCL2 file produces an identical dict when parsed, even after reconstruction
  • test_write_terraform_exact ensures that the reconstructed HCL2 exactly matches the input HCL2, down to the individual character and whitespace level.
    • Note: some files are hardcoded within this test method to be skipped, as the reconstruction doesn't perfectly reproduce them in its output. This is OK, because the output is still identical when parsed. This behaviour is verified by the above test_write_terraform test, which has no such exceptions.

I have verified that all existing tests are passing, and overall test coverage is improved by this PR. Please let me know if you have any questions, or suggested changes.

[kkozik-amplify]

@weaversam8 Hi, thank you for contributing! The PR looks good - we will release this today with 5.0.0 tag