perf(concepts): parallelize /build/concepts link-table reads

[jung-thomas]

Problem

/build/concepts takes ~50–60s on DEV — past the 30s smoke-gate timeout. Surfaced as 4 smoke failures (all pure 30s timeouts, not 4xx/5xx) in the DEV 1.25.0 deploy: concepts-page, concepts-route, asset-retention, explore-route. The endpoint returns correct 200s, just slowly.

Not a 1.25.0 regression — published-concepts-query.js last changed 2026-08-09; the slowness is structural.

Root cause

buildConceptsPayload issued its ~11 independent link-table reads (teaches, edges, journeys, blogs, missions, videos, apiDocs, samples, helpDocs, events) strictly sequentially — each awaited before the next. Each is an unbounded full-table scan filtered in Node (deliberate, to dodge the HANA IN (ids) packet-size bug, #1103). On HANA that's ~11 serial round-trips over multi-thousand-row tables per cold call.

Fix

The reads are all keyed off idsSet (computed once in step 1) with no inter-query dependencies → wrap them in one Promise.all wave. Steps 3+4 stay grouped (they jointly build requires/relatedTo/requiredBy) and run their two edge reads in parallel too. Stitch (step 6) and groupBy unchanged; identical columns, filters, sorts, per-concept caps.

Impact / risk

• No user impact was ever in play — /concepts/* pages are served from HANA BLOBs (fast, verified 0.8s). This endpoint feeds only the Hugo concept-render build pipeline and the smoke gate.
• Behavior identical: same reads, same output shape. All 744 concepts-related unit tests pass.
• Concurrency note: ~12 in-flight db.run may briefly exceed the HANA pool (default max 10); excess queries queue rather than fail — still overlaps, no correctness risk.

Verification

• npx vitest run --project unit over all concepts suites → 744 passed, 1 skipped.
• Recommend confirming wall-clock on DEV post-merge (expect well under the 30s smoke bound); the 4 smoke timeouts should clear.