feat: Allow specifying the base commit for local (non-CI) diffs (#180)

* Add option to specify base commit
* Add test
* Update docs

---------

Co-authored-by: Brendan <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>

fix: skip parsing clang-tidy diagnostic rationale (#176)

Specifically when the rationale line has added context in square brackets.

This includes a test based on the actual bug report in cpp-linter/cpp-linter-action#389.

feat: support python 3.14 (#167)

fix: prevent duplicate files when `--files-changed-only=false` (#164)

I found that the glob pattern for some extensions hit multiple files more than once.

```py
pathlib.Path(".").rglob("*.c")
# matches demo.c and demo.cpp
```

Worse, the duplicates were being analyzed more than once.
So, this offer a performance improvement as well.

fix: parse clang-tidy output when WarningsAsErrors is asserted (#162)

ref cpp-linter/cpp-linter-action#347

fix: specify stale workflow permissions for org reusable workflow (#160)

---------

Co-authored-by: copilot-swe-agent[bot] <[email protected]>
Co-authored-by: shenxianpeng <[email protected]>
Co-authored-by: Brendan <[email protected]>

Replace deprecated classifier with licence expression (PEP 639) (#136)

add: File IO timeout API (#133)

Born from the discussion in #129 and continued in #130. This adds timeout-oriented functions for reading and writing to the scanned source files.

This supersedes #130 and resolves #129.

* run clang-format before clang-tidy
* prevent busy waits and log errors

prefix every review comment (#132)

With the html marker used in our other generated comments. This is to avoid ambiguity and help on-board new features in #131.

fix: enhance parsing paginated diffs (#125)

- fixes libgit2 diff parsing errors in paginated responses about changed files' diff.
- allows renamed files to be scanned entirely when the source file's content has not changed (when using paginated responses)
- add tests, but ignores coverage for critical errors (like malformed JSON responses)