fix: keep issue discovery scoring when open-count fetch fails

[Matsx901]

Summary

Closes #1249.

Issue discovery currently restores cached issue-discovery fields and skips a miner when the secondary open-issue count fetch fails, even if the primary lookback fetch already succeeded. This discards fresh solved-issue evidence because the open-count request is treated as a hard dependency.

This change keeps the successful scoring response and only falls back the open-count input.

Root Cause

run_issue_discovery() performs two mirror requests per miner:

• lookback-bounded request for scoring data
• current request for open-issue spam count

The second request failure path restored cached issue-discovery fields and continued, so fresh solved/valid-solved data from the first request was ignored.

Changes

• Keep first-fetch failures as hard fallback to cached issue-discovery fields.
• Treat second/open-count fetch failures as soft failures.
• Use cached total_open_issues when available.
• Fall back to counting open issues from the already-fetched lookback response when no cache is available.
• Add regression coverage for cached and no-cache fallback paths.

Out of scope

• No changes to issue eligibility rules.
• No changes to open-issue spam multiplier thresholds.
• No changes to solving PR scoring or cache-miss behavior.

Type of Change

• Bug fix
• New feature
• Refactor
• Documentation
• Other

Testing

uv run pytest tests/validator/issue_discovery/test_scan.py -q
uv run ruff check .
uv run ruff format --check .
uv run pyright
uv run pytest -q
git diff --check

[anderdc]

Closing — see #1251 for the documented copying of @seroperson's vulture-integration branch (14 of 14 touched files a subset of the source PR, 12 byte-identical, and the linked issue body explicitly cites the source branch by name). With that on the public record, we can't extend the good-faith review this PR would otherwise need.

Closing.