Clean testing on a clean up test corpus

[dsimunic]

Resolved all real failures in the test. A few real failures were linked to multiline string parsing: after replacing the regex with a new token in the scanner, all parses are correct.

While at it, also resolved the bug with nesting block comments. Closes #78

About 50 files were elm 0.18 or syntax errors. Checked each one individually on Elm 0.19 compiler. Blacklisted those files in expected_fails.txt. The existing parse-examples-full will probably report 51 error since it doesn't know about the blacklist.

Finally, I cleaned up the test corpus: instead of downloading each repo and risking that something changes, I moved all .elm files and licenses into a clean repo. This gives us a stable test inputs. I've added a new script parse-test-corpus that works on that corpus, and the corresponding workflow action test-on-test-corpus.yml.

$ script/parse-test-corpus --expect-fail script/expected_fails.txt  /Volumes/Devel/var/tree-sitter-elm/tree-sitter-elm-test-corpus
Parsing...
-----------------------------------------------------------------
No unexpected failures!

All expected failures failed as expected.

Summary:
  Total files parsed: 20119
  Successful parses: 20119
  Unexpected failures: 0
  Unexpected successes: 0
  Total errors: 0
  Success rate: 100.00%
-----------------------------------------------------------------

Took: 17.253 seconds

If you agree with that new workflow, it obsoletes a bunch of test scripts that I can delete in another PR.

The test files are presently in my GitHub account, but I'd like to move that repo over to elm-tooling if you're ok with that.

[dsimunic]

For whatever reason, the env var didn't pick up the default. For the new testing to pass we need to fix the path to the test corpus repo in repo vars.

CORPUS_PATH: ${{ vars.TEST_CORPUS_PATH || 'https://github.com/dsimunic/tree-sitter-elm-test-corpus' }}

Alternatively I can hard-code the path and then we fix it in another PR if you agree to move the data over to elm-tooling.

[dsimunic]

Hey! Is there anything else you expect from me on this one?

[razzeee]

Looking at this, I don't really like the test changes.

If you really would want the tests to not run against head, just change the old script to checkout the tags instead of head?