fix(release): drain v3.8.50 base-reds — build-breaking import, stale provider docs, orphaned suppressions

[diegosouzapw]

Refs #9985

Drains three classes of base-red currently failing every PR against release/v3.8.50 (complementary to the in-flight compression basereds branch — zero file overlap):

1. Fast Production Build (module-not-found) — src/app/api/usage/utilization/route.ts imported getConnection from @/lib/db/connections, a module that does not exist (introduced by the fix(dashboard): show account email/name in Utilization Account Split cards #10939 merge, c0fd109; typecheck:core does not cover app routes, so only next build catches it). Fixed to getProviderConnectionById from @/lib/db/providers (awaited). Regression test: tests/unit/utilization-route-import-10939.test.ts (module import is the repro — ERR_MODULE_NOT_FOUND on the base, resolves here).
2. Docs Gates (fast-path) — docs/reference/PROVIDER_REFERENCE.md regenerated via npm run gen:provider-reference (346 → 347) and the four README diagram SVGs updated to the real count.
3. No new ESLint warnings — npm run lint:json -- --max-warnings 0 --prune-suppressions on the clean base tip removed suppressions orphaned by earlier merges (the gate fails with "suppressions left that do not occur anymore" for every PR); lint exits 0 after the prune.

[diegosouzapw]

Thanks for chasing this down — the work here is sound and the intent (drain the #10939 build-breaking import) is exactly right. However, this PR has been superseded: the same fix (getProviderConnectionById from @/lib/db/providers) already landed on release/v3.8.50 via commit 137e49e (#10988), which is now on the release tip, and a stronger regression test (tests/unit/usage-utilization-connection-meta.test.ts) already ships there covering the phantom-import guard plus a full GET/connection-meta run. Your head is 25 commits behind the current tip, so the ESLint-suppression pruning and docs regenerated to a "347" count rest on a stale base (the tip's README/PROVIDER_REFERENCE already read 348). I'd recommend against force-merging this stale branch: it would conflict with already-correct code and could regress refreshed docs. Better path: verify the current tip's utilization route + its existing connection-meta test are green (the base-red for #10939 is no longer listed among current release reds), and close this PR without merging rather than re-apply the patch — [OPTIONAL] if the 685 suppression lines and provider-count docs are genuinely stale, re-cut a fresh, small cleanup PR off the current tip. [MANDATORY — nothing to merge here; the fix is already integrated.]