fix: require solved_by_pr match for mirror linked-issue multiplier

[carlos4s]

Summary

_is_valid_linked_issue walked every anti-gaming gate for the issue multiplier — transferred, self-issue, created_at, state_reason, edited_after_merge, close-window — except solver attribution. The field MirrorLinkedIssue.solved_by_pr was already populated by das-github-mirror and treated as authoritative in issue_discovery/scan.py (scan.py:3,13), but the PR-multiplier path ignored it. A PR that referenced an issue but was not the actual solver could still receive STANDARD_ISSUE_MULTIPLIER or MAINTAINER_ISSUE_MULTIPLIER whenever the other gates happened to pass.

This PR adds a single guard alongside the other PR-state-agnostic gates so both surfaces are covered:

• Merged PRs within the +0..+1 day close-window after another PR closed the linked issue (the close-window gate passes; nothing else caught the wrong-solver case).
• Open PRs on the collateral path, which skips the close-window gate entirely and previously had no solver-attribution check at all.

The gate fails open when solved_by_pr is None to preserve behavior on older mirror snapshots that don't populate the field. This matches the conservative posture used elsewhere in the validator.

One-PR-per-issue semantics

If multiple PRs reference the same issue, only the one matching the mirror's solved_by_pr keeps the issue multiplier; the rest fall back to 1.0× on the issue slot (they still earn their full base_score). This mirrors issue_discovery, which already awards discovery_earned_score to a single solver per issue. Loosening to "any PR that referenced the issue" would re-open the multiplier-farming gap this PR closes.

Mirror attribution

If the mirror's solved_by_pr is wrong for a given issue, that's a mirror-side correction, not a validator-side relaxation. The validator's job is to consume the field consistently across paths.

Related Issues

Close #1353

Type of Change

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

Testing

• Tests added/updated
• Manually tested

New test class TestLinkedIssueSolvedByPr in tests/validator/oss_contributions/mirror/test_scoring.py covers:

• merged PR with mismatched solved_by_pr → rejected
• open PR with mismatched solved_by_pr (collateral path) → rejected
• matching solved_by_pr → still valid (regression guard)
• solved_by_pr is None → still valid (fail-open regression guard)
• end-to-end _calculate_issue_multiplier returns 1.0 on mismatch
• maintainer-authored mismatch still rejected (no MAINTAINER_ISSUE_MULTIPLIER shortcut)

Commands run locally on the changed files and the full suite:

• uv run pytest tests/validator/oss_contributions/mirror/test_scoring.py -q → 67/67
• uv run pytest tests/ -q → 864/864
• uv run ruff check gittensor/ tests/ → clean
• uv run ruff format --check gittensor/ tests/ → clean
• uv run pyright gittensor/validator/oss_contributions/mirror/scoring.py tests/validator/oss_contributions/mirror/test_scoring.py → 0 errors / 0 warnings
• uv run pre-commit run --hook-stage pre-push --files <changed files> → all hooks pass (vulture, pyright, pytest)

Checklist

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