fix: use SOURCE-quality score for eligibility gates

[bitloi]

Summary

Eligibility-style gates were using aggregate token_score, which includes tree-diff TEST score, even though the base-score threshold already uses SOURCE-only quality. A large test-only PR could therefore get only the intended small contribution bonus, but still count as a valid merged PR, valid solved issue-discovery PR, or pioneer-eligible PR.

This change:

• Persists source_token_score from calculate_base_score_for_pr_files().
• Keeps aggregate token_score unchanged for contribution accounting and token totals.
• Uses SOURCE-quality score for merged PR eligibility, mirror valid-solved gates, and legacy/mirror pioneer eligibility.
• Falls back to aggregate token_score when source_token_score is absent so older/test objects remain compatible.
• Adds regression coverage showing TEST-only PRs can earn the small bonus without satisfying SOURCE-quality gates.

Related Issues

Fixes #1022.

Related context:

• Tests affect code density and reduce resulting score very much #339
• Scoring tests separately #350

Type of Change

• Bug fix
• New feature
• Refactor
• Documentation
• Other (describe below)

Testing

• Tests added/updated
• Manually tested

Commands run:

git diff --check
uv run pytest tests/validator/test_source_quality_gates.py tests/validator/issue_discovery/test_mirror_scan.py tests/validator/oss_contributions/mirror/test_base_score_helper.py tests/validator/oss_contributions/mirror/test_scored_pr.py tests/validator/test_pioneer_dividend.py -q
uv run ruff check gittensor/classes.py gittensor/constants.py gittensor/validator/issue_discovery/mirror_scan.py gittensor/validator/issue_discovery/scoring.py gittensor/validator/oss_contributions/credibility.py gittensor/validator/oss_contributions/mirror/adapters.py gittensor/validator/oss_contributions/mirror/scored_pr.py gittensor/validator/oss_contributions/mirror/scoring.py gittensor/validator/oss_contributions/scoring.py tests/validator/issue_discovery/test_mirror_scan.py tests/validator/oss_contributions/mirror/test_base_score_helper.py tests/validator/oss_contributions/mirror/test_scored_pr.py tests/validator/test_pioneer_dividend.py tests/validator/test_source_quality_gates.py
uv run ruff format --check gittensor/classes.py gittensor/constants.py gittensor/validator/issue_discovery/mirror_scan.py gittensor/validator/issue_discovery/scoring.py gittensor/validator/oss_contributions/credibility.py gittensor/validator/oss_contributions/mirror/adapters.py gittensor/validator/oss_contributions/mirror/scored_pr.py gittensor/validator/oss_contributions/mirror/scoring.py gittensor/validator/oss_contributions/scoring.py tests/validator/issue_discovery/test_mirror_scan.py tests/validator/oss_contributions/mirror/test_base_score_helper.py tests/validator/oss_contributions/mirror/test_scored_pr.py tests/validator/test_pioneer_dividend.py tests/validator/test_source_quality_gates.py
uv run pyright gittensor/classes.py gittensor/validator/issue_discovery/mirror_scan.py gittensor/validator/oss_contributions/credibility.py gittensor/validator/oss_contributions/mirror/scored_pr.py gittensor/validator/oss_contributions/mirror/scoring.py gittensor/validator/oss_contributions/scoring.py tests/validator/test_source_quality_gates.py
uv run pytest tests/ -q

Results:

• Focused regression suite: 68 passed
• Full test suite: 769 passed, 2 warnings
• Ruff check: passed
• Ruff format check: passed
• Pyright on touched files: 0 errors, 0 warnings

Checklist

• Code follows project style guidelines
• Self-review completed
• Changes are documented (if applicable)

[anderdc]

Maintainer review already rejects test-only PRs, so reaching the failure mode this defends against — five test-only PRs accumulating aggregate token_score ≥ 5 from the 0.05× TEST weight — requires maintainer cooperation that doesn't happen in practice. Not pursuing the fix. Closing.