Add an option to skip leading context lines for a more correct parse #160
Description
The from-repo subcommand of the diff-annotate script supports the --use-repo option,
which makes the script retrieve pre-/post-image contents from the repository, and use it for lexing with Pygments,
instead of using the post-image excerpt (fragment) of the code extracted from the patch / changeset.
The goal of that option is to avoid mis-parse (by Pygments), and mis-attribution of lines,
if the code fragment (post-image extracted from patch) starts in the middle of the code constructs.
For example:
- multi-line comment in C, C++, Java
- docstring in Python, Bash, Kotlin
This approach is not possible for when we don't have the repo, or don't have the information about the commit
(or pre-/post-image contents, via extended diff header, like for example index f4b1722757..a3765ae0f1 100644).
Add an option which makes the diff-annotate script to try to parse reconstructed pre- and post-image,
with trimmed leading context lines - as for annotation we need to have correct parse only for added and deleted lines.
This can be done with some error-resistant parser, like parso, by finding the parse with least percentage of errors.