fix: exclude deleted files from code_density calculation

[t7929375-eng]

Summary

When a PR deletes files, the deleted files are correctly scored at 0 (scoring_method='skipped') in tree_sitter_scoring.py, but their total_lines (set to file.deletions) are still included in the total_lines sum in scoring.py:172.

Since code_density = token_score / total_lines, deleted files inflate the denominator without contributing to the numerator, artificially deflating code_density and reducing the base_score for all miners whose PRs include file deletions.

This fix filters out skipped (deleted) files from the total_lines calculation so that code_density accurately reflects only the scored content.

Example: A PR with 50 lines of dense code (token_score=80) that also removes a 200-line file would get code_density = 80/250 = 0.32 instead of the correct 80/50 = 1.6 — a 5x penalty.

Type of Change

• Bug fix

Testing

• ruff check passes
• pyright passes
• Self-reviewed

Checklist

• Code follows project style guidelines
• Self-review completed

cc @anderdc @LandynDev