feat(cache): serve stale metrics on GitHub rate limits

[bhavyajain0810]

Summary

Adds stale-cache fallback support for GitHub-backed metrics so previously successful data can still be returned when GitHub temporarily rate-limits a refresh. The pinned repositories endpoint now uses the shared cache and centralized GitHub GraphQL handling while preserving authentication and non-rate-limit error behavior.

Closes #1930

---

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 💥 Breaking change (fix or feature that changes existing behavior)
• 📝 Documentation update
• ♻️ Refactor / code cleanup (no functional change)
• ⚡ Performance improvement
• 🔒 Security fix
• 🧪 Tests only

---

What Changed

• Extended src/lib/metrics-cache.ts with an optional stale-cache layer, configurable stale-grace period, and error predicate for controlled fallback.
• Updated src/app/api/metrics/pinned-repos/route.ts to use withMetricsCache, cache-bypass handling, and the centralized githubGraphQL helper.
• Added fallback behavior that serves stale pinned-repository data only when a GitHub rate-limit error prevents fresh data from loading.
• Updated cache deletion so both the fresh entry and its corresponding stale entry are removed.
• Added tests in test/metrics-cache.test.ts for stale fallback, unrelated errors, explicit cache bypass, and deletion of stale entries.

---

How to Test

1. Run the focused lint checks:

   npx eslint src/lib/metrics-cache.ts src/app/api/metrics/pinned-repos/route.ts test/metrics-cache.test.ts test/pinned-repos.test.ts

2. Run the relevant tests:

   npm test -- test/metrics-cache.test.ts test/pinned-repos.test.ts

3. Run TypeScript validation:

   npm run type-check

Expected result:

• ESLint completes without errors.
• Both relevant test files pass: 23 tests passed.
• tsc --noEmit completes without TypeScript errors.
• A normal cache hit avoids a GitHub request.
• After the fresh entry expires, a recognized GitHub rate-limit error returns the stale value.
• Unrelated errors are still propagated instead of being hidden by stale data.
• Explicit cache-bypass requests do not use fresh or stale cache fallback.

---

Checklist

• Linked the related issue above
• Self-reviewed my own diff
• No unnecessary console.log, debug code, or commented-out blocks
• npm run lint passes locally
• No TypeScript errors (npm run type-check)
• Added or updated tests where applicable
• Updated documentation / comments if behavior changed

---

Additional Context

• This is an internal API and caching change with no UI impact, so screenshots and accessibility testing are not applicable.
• Stale fallback is opt-in and only activates when the supplied error predicate approves the failure.
• Cache-bypass requests continue to force a fresh request and do not silently fall back to stale data.
• Existing authentication failures and unrelated GitHub API failures continue to return their normal error responses.

[github-actions]

GSSoC Label Checklist 🏷️

@Umbrella-io — please apply the appropriate labels before merging:

Difficulty (pick one):

• level:beginner — 20 pts
• level:intermediate — 35 pts
• level:advanced — 55 pts
• level:critical — 80 pts

Quality (optional):

• quality:clean — ×1.2 multiplier
• quality:exceptional — ×1.5 multiplier

Validation (required to score):

• gssoc:approved — counts for points
• gssoc:invalid / gssoc:spam / gssoc:ai-slop — does not score

Type labels (type:*) are auto-detected from files and title. Review and adjust if needed.
Points formula: (difficulty × quality_multiplier) + type_bonus

[bhavyajain0810]

Hi @Priyanshu-byte-coder,

CI status note

The two failing Playwright checks appear unrelated to the changes in this PR.

E2E smoke test

The failing test is:

e2e/streak.spec.ts
[Streak E2E] streak widget shows the mocked longest streak value

Playwright reports a strict-mode violation because:

[aria-label="Your longest streak ever"]

matches two existing elements: the longest-streak card and its information button. The same failure occurred across retries, while the other 46 smoke tests passed.

Visual regression

The visual-regression job failed in these existing UI scenarios:

landing page full page screenshot in dark mode
public profile screenshot with deterministic mock data

The public-profile test timed out while waiting for the expected profile heading. Neither these pages nor their visual tests are modified by this PR.

This PR only changes:

• src/lib/metrics-cache.ts
• src/app/api/metrics/pinned-repos/route.ts
• test/metrics-cache.test.ts

All relevant validation is passing, including build, lint, type-check, test coverage, dependency audit, environment security checks, and the focused cache and pinned-repository tests.

Could you please review or rerun these unrelated Playwright failures?