fix: count old open mirror issues for issue spam multiplier

[MkDev11]

Summary

Splits mirror issue-discovery scoring from open-issue spam counting.

The scoring path still uses the lookback-bounded mirror /issues response, but total_open_issues is now resolved independently so older still-open issues count toward calculate_open_issue_spam_multiplier().

This also adds support for a future mirror open_issue_count response field. Until the mirror serves that field, the validator falls back to a separate all-time current-open issue fetch and filters it to mirror-enabled repos.

Related Issues

Fixes #929

Type of Change

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

Testing

• Tests added/updated
• Manually tested

Commands run:

uv run pytest tests/validator/issue_discovery/test_mirror_scan.py tests/utils/test_mirror_models.py tests/utils/test_mirror_client.py
uv run ruff check gittensor/utils/mirror/models.py gittensor/validator/issue_discovery/mirror_scan.py tests/validator/issue_discovery/test_mirror_scan.py tests/utils/test_mirror_models.py
uv run ruff format --check gittensor/utils/mirror/models.py gittensor/validator/issue_discovery/mirror_scan.py tests/validator/issue_discovery/test_mirror_scan.py tests/utils/test_mirror_models.py
uv run pyright
git diff --check

Checklist

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

[anderdc]

Still debating on merging this — putting a pin in it.

[MkDev11]

thanks @anderdc , totally understand you.

My only push for merging is that this is the exact follow-up called out in #796: scoring should stay lookback-bounded, but the open-issue spam count should be current open issue load.

The PR keeps the scoring fetch unchanged. It only separates total_open_issues, prefers the future open_issue_count field when present, and falls back to current behavior if the extra count fetch fails.

If the concern is the temporary second mirror request, I can drop that fallback and only wire the future field, but that would leave #929 unfixed until the mirror serves it.

[anderdc]

das-github-mirror awaiting push to prod so this is done in a simplified manner. Please be on standby for further direction on this PR in the next few days.

[anderdc]

das-github-mirror's since-omitted issues endpoint is now in prod (entrius/das-github-mirror#27). When the issues request omits since, the response is all currently-OPEN issues across mirror-tracked repos, no CLOSED rows. That makes both compatibility paths in this PR obsolete — drop them and call the endpoint with no since directly.

Specifics:

• gittensor/utils/mirror/models.py — remove the open_issue_count field on MirrorIssuesResponse and its parsing block. It won't be served; the no-since request shape replaced it.
• gittensor/validator/issue_discovery/mirror_scan.py — replace _resolve_current_open_issue_count with a single client.get_miner_issues(github_id) call (no since). The result is _count_open_issues(current_response.issues, enabled_names). Drop _OPEN_ISSUE_COUNT_SINCE, the open_issue_count is not None short-circuit, and the MirrorRequestError fallback to the bounded response — a failed open-count fetch should take the same fetch_errors += 1; continue path the scoring fetch already uses, not silently fall back to the wrong signal.
• tests/utils/test_mirror_models.py — drop the two open_issue_count tests.
• tests/validator/issue_discovery/test_mirror_scan.py — delete test_open_issue_count_field_avoids_second_fetch. Keep test_old_open_issues_outside_scoring_window_still_trip_spam but assert the second get_miner_issues call passes since=None (or no since kwarg), not _OPEN_ISSUE_COUNT_SINCE.

Also rebase on test to clear the conflict.