Extend infinite scroll to portal list surfaces (Closes #972)

[cjimti]

Summary

Closes #972. The Assets library loads all items via infinite scroll, but the other portal list surfaces fetched only their first page, so a user or deployment with more items than one page silently could not reach the rest. This PR extends the same pattern to every affected surface that has a stable, offset-capable backend, so each list can display its full contents.

Surfaces now paginated

Each of these accumulates pages as you scroll, with a "Load more" button as the fallback and indicator, matching the Assets library:

• Collections (My Collections and Shared)
• Feedback: Recent (activity feed), Worklist (practitioner and SME), and General (standalone channel)
• Resources (My Resources, Global, and per-persona/admin tabs)
• Knowledge Pages
• Settings > Users
• Settings > Changelog

Shared infrastructure

Paging is implemented once and reused, rather than reimplemented per page:

• ui/src/api/portal/hooks/infinite.ts holds useOffsetInfiniteQuery plus the pure helpers nextOffset, flattenPages, toPaginated, and paginatedFetch. The Assets hooks now consume this shared module too (the paging helpers moved out of assets.ts, which re-exports them for backward compatibility).
• ui/src/components/InfiniteFooter.tsx is the shared Load-more control: it owns the scroll sentinel and wires useInfiniteScroll, and renders nothing once every page is loaded.
• Rows are merged in fetch order and de-duplicated by a stable key, so offset paging over a created_at/updated_at DESC list cannot double-count a row when an insert shifts the window. Paging stops on the latest page's total or an empty trailing page, so "Load more" cannot spin forever.
• Endpoints that return a non-PaginatedResponse envelope (Resources {resources,total}, Users {users,total}, Knowledge Pages {pages,total}, Changelog {entries,total}) adapt their response to the shared shape at the fetch boundary via toPaginated, so all surfaces share one merge, dedupe, and stop-condition path.

Backend

• Resources: handleList now honors a client-supplied limit, clamped to a new resource.MaxListLimit. Previously it ignored the client limit and always used the default page size.
• Config changelog: configstore.Store.Changelog gained offset and a full-history total (Postgres OFFSET plus a COUNT(*), the file store, and the admin handler), returned as a {entries,total} envelope and clamped to maxChangelogLimit. Older entries beyond the most-recent window are now reachable.
• Swagger annotations and generated internal/apidocs are updated for the new limit parameter and the changelog response shape.

DataHub Catalog and Context Docs (not wired)

Catalog and Context Docs browse are intentionally left single-page. They page against an external DataHub via a query:"*" search that sends no sort criteria, so results come back in relevance/segment order that is not stable across requests: numeric offset paging over it can drop or duplicate rows between fetches, and catalog browse returns no total. Infinite scroll here would be a correctness regression worse than the current cap, so it stays gated on the upstream mcp-datahub client gaining a stable sort or a scroll/search-after cursor. The constraint is documented in ui/src/api/portal/datahub.ts. Search remains the escape hatch.

Tests and verification

• New shared hooks and the InfiniteFooter control have unit tests; the feedback component tests are updated to the infinite hooks.
• dev/seed.sql seeds more than one page of resources, collections, threads, knowledge pages, users, and changelog entries so every surface exercises pagination in the dev environment. The seed is validated against a real Postgres by make migrate-check.
• MSW mocks for the changelog and resources endpoints window by limit/offset so the e2e and screenshot pipelines exercise paging.
• Full make verify passes locally (unit tests with race, coverage and patch coverage, golangci-lint, gosec, govulncheck, semgrep, CodeQL, frontend tests and e2e, swagger-check, migrate-check).

Notes

• The config changelog request is clamped to a maximum page size, matching the clamp every other paginated store applies, so a client cannot request an unbounded window.
• The "Showing X of Y" caption on paginated lists is shown only while more pages can be loaded, so a de-duplicated count never leaves a permanent off-by-one with no control to resolve it.
• Knowledge Pages request 100 per page, the store's largest honored search limit; the previous fixed limit: 200 request was silently collapsed to 20 by the store, so the tag facet is now more complete on first load, not less.

Follow-up (out of scope)

Real pagination for the currently unbounded "fetch-all" endpoints (My Prompts, Admin Prompts, Prompt Review Queue, Tools, and the Settings lists) remains a lower-priority follow-up, and DataHub-backed browse is blocked on the upstream paging guarantee described above.

[codecov]

Codecov Report

❌ Patch coverage is 93.10345% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.06%. Comparing base (01d36b1) to head (97af12c).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/configstore/postgres/store.go	83.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #974      +/-   ##
==========================================
+ Coverage   89.03%   89.06%   +0.02%     
==========================================
  Files         424      424              
  Lines       48581    48598      +17     
==========================================
+ Hits        43256    43283      +27     
+ Misses       3549     3543       -6     
+ Partials     1776     1772       -4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.