fix: gate retroactive issue body edits in mirror issue discovery

[statxc]

Summary

Restores legacy parity in mirror_scan._classify_issue. PR #796 ported the PR-side post-merge edit gate (solving_pr.edited_after_merge at mirror_scan.py:461) but missed the issue-side counterpart (issue.body_or_title_edited_at > pr.merged_at, originally added in #434). MirrorIssue.last_edited_at is already fetched and parsed at gittensor/utils/mirror/models.py:271 — it just was never read by the scoring path.

Without this gate, a miner can author a deliberately vague issue, wait for a third party's PR to merge a fix, then edit the issue body to retroactively match what the PR fixed and collect full discovery_earned_score they didn't earn. The check is added inline in _classify_issue, mirroring the shape of the neighboring solving_pr.edited_after_merge block. Returns 'not-solved-closed' so the issue still counts toward credibility (matches legacy semantics) but contributes no discovery score.

Legacy parity table

Gate	Pre-#796	Post-#796	This PR
PR body edited after merge	oss_contributions/scoring.py:493 checked pr.last_edited_at > pr.merged_at	✅ Ported as solving_pr.edited_after_merge	unchanged
Issue body/title edited after merge	issue_discovery/scoring.py:253 checked issue.body_or_title_edited_at > pr.merged_at (added by #434)	❌ No replacement	✅ Ported as issue.last_edited_at > sp.merged_at

Relationship to #822

#822 fixes the cross-miner one-issue-per-PR rule in run_mirror_issue_discovery. This PR fixes a missing anti-gaming gate in _classify_issue. Different functions, different test classes, different attack vectors — the only mechanical overlap is the _issue_dict test helper signature, which 3-way merges cleanly in either order.

Related Issues

Closes #826

Type of Change

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

Testing

• Tests added/updated
• Manually tested

Three new tests in TestClassifyIssue (mirroring the existing test_solving_pr_edited_after_merge_counts_as_closed pattern):

Test	Setup	Asserts
test_issue_edited_after_solving_pr_merge_counts_as_closed	last_edited_at = merged_at + 1s	'not-solved-closed' (gate fires)
test_issue_edited_before_solving_pr_merge_is_solved	last_edited_at = merged_at − 1d	'solved' (legitimate pre-merge edits not penalized)
test_issue_never_edited_is_solved	last_edited_at = None	'solved' (gate is null-safe)

Commands run:

• uv run ruff check gittensor/validator/issue_discovery/mirror_scan.py tests/validator/issue_discovery/test_mirror_scan.py — All checks passed
• uv run ruff format --check <same paths> — 2 files already formatted
• uv run pyright <same paths> — 0 errors, 0 warnings, 0 informations
• uv run python -m pytest tests/validator/issue_discovery/test_mirror_scan.py -v — 34 passed (3 new + 31 existing)
• uv run python -m pytest (full suite) — 646 passed

Out of scope

• The PR-body gate (solving_pr.edited_after_merge) is correctly ported and unchanged here.
• This PR uses the mirror's authoritative last_edited_at directly — it does not populate the legacy body_or_title_edited_at field on the adapted Issue, since that field is unread on the mirror path (mirror_scan.py:504 already passes body_or_title_edited_at=None).

Checklist

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