Handle ast.parse failures when converting function type comments

[stroxler]

Summary

We have to parse code using the ast module in two situations, and in either case we can hit parse errors even though the code is valid python.

(a) by parsing with type_comments=True to reliably extract string type comments from code. This can throw errors if there are comments of the form # type: something anywhere where type information is invalid according to PEP 484.

(b) by parsing the type comments themselves into expressions or ast.FunctionType values. This can throw errors, even if (a) succeeds, whenever the type itself is not well-formed.

This commit updates convert_type_comments to handle these situations by ignoring type information on a specific node - prior to this commit, the entire codemod would have crashed, but there may be many other type comments in the module which are perfectly valid and we'd like to be able to keep going when we hit SyntaxErrors.

Test Plan

I added new test cases for invalid type comments

[codecov-commenter]

Codecov Report

Merging #616 (1d620da) into main (ceac7fc) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main     #616   +/-   ##
=======================================
  Coverage   94.78%   94.78%           
=======================================
  Files         245      245           
  Lines       25238    25250   +12     
=======================================
+ Hits        23921    23934   +13     
+ Misses       1317     1316    -1     

Impacted Files	Coverage Δ
libcst/codemod/commands/convert_type_comments.py	95.98% <100.00%> (+0.69%)	⬆️
...demod/commands/tests/test_convert_type_comments.py	96.72% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ceac7fc...1d620da. Read the comment docs.