Add spell checker to github workflows

[reyammer]

The PR #1240 found quite a lot of typos.

This is the command they used: codespell -S *.json,*.lock,tests_data -L hve,afterall and typos --hidden --format brief.

Check whether we can automate something like this (or at least raise warnings). Unclear what's the false positives rate.

We can either extend the scope of https://github.com/google/magika/blob/main/.github/workflows/docs-check.yml to check for more than documentation, or add a new spell checker workflow across the entire codebase.

[ps2181]

Hi @reyammer

A few ideas to keep the signal high while limiting overhead for this:

• Warning-only mode: We can set the workflow to continue-on-error: true initially. This lets us observe false positives in the logs without blocking PR merges.

• Targeted Scoping: Focus on high-value areas first:

    -docs/**, *.md, *.rst
  
    -python/** (specifically excluding tests_data, .json, .lock, and generated files).
  

• Centralized Allowlist: Instead of cluttered inline -L flags in the YAML, we could use a .codespellrc file. This makes it easier for contributors to document and justify domain-specific terms as they arise.

• Differential Checks: We can configure the action to run only on modified files in a PR. This prevents "noise" from legacy typos and keeps CI runs fast.

• Strict Exclusions: Maintain the current exclusions for problematic paths (*.json, *.lock, tests_data) to avoid scanning binaries or data fixtures.

@reyammer do let me know how if this approach works?