Add Company and Job Title filters to Members page

[Bjjj834]

Problem

The Members page (/community/members, ~993 members / ~100 pages) only had a single free-text search box. Users had no way to narrow the list by Company or Job Title specifically.

Current Members page behavior (investigated before making changes)

• Rendered by ArmorMembers.js, routed at /members under Main.js.
• Data source: the browser fetches the entire Google Sheet (SHEET_ID=1IREd2vCxo7rDGOoDUYtzoywVGLqqUFCsu3qGsA0HsMc, tab Sheet-1) directly from the Google Sheets API in one request — no backend, no database, no paginated API.
• Pagination is client-side only (filtered.slice((page-1)*pageSize, page*pageSize), pageSize=10) over the full in-memory array.
• The existing search already filtered the full loaded dataset, not just the current page — this was correct and preserved.
• Field names: Company Name column → member.company, Job Title column → member.title.
• Conclusion: since the full dataset is already safely loaded client-side, the new filters could be implemented entirely in this repo with no backend dependency and no risk of only filtering the visible page. There was no external-backend blocker.

Proposed solution

• memberFilters.js — pure, unit-tested helpers: value normalization, deduped/sorted option lists (blank-free), and the three match functions (general search, Company = exact case/whitespace-insensitive match, Job Title = case-insensitive keyword/contains match).
• FilterCombobox.js — a shared searchable dropdown/autocomplete used for both filters, with keyboard navigation, an individual clear button, and accessible labeling.
• ArmorMembers.js — wires both filters in next to the existing search box, combines all three conditions with AND semantics, adds removable filter chips, a "Clear filters" action, and an empty state ("No members match the selected filters.") with its own clear action.
• useDebouncedValue.js — 250ms debounce for the free-text search box.
• setupTests.js — @testing-library/jest-dom was already a devDependency but never wired up (no setupTests.js existed), so matchers like toBeInTheDocument/toHaveValue weren't available to any test in the suite. Added the standard CRA setup file.

Data and API changes

None. No backend/API changes — everything operates on the dataset already loaded into the browser, per the investigation above.

Filter and pagination behavior

• Filters combine with the general search using AND semantics.
• Changing any filter resets pagination to page 1 (skipped on initial mount so a shared/reloaded URL keeps its page).
• Page is clamped into range once the filtered dataset is known (handles a shared URL like ?page=999 against a narrower filter).
• Filters persist across Next/Previous page navigation.
• Search box is debounced (250ms); Company/Job Title commit on selection, Enter, or blur (not per keystroke), so there's no per-keystroke re-filtering to debounce there.
• Search/company/jobTitle/page are synced to the URL (e.g. ?company=Anote&jobTitle=CEO&page=2), empty params omitted, so the page is refresh-safe and shareable.

Manual verification (real data, not mocks)

Ran the app locally against the live Google Sheet (993 real members, 100 pages) and verified:

• Company "Anote" → 132 real matches across 14 pages (was previously unreachable via the old 10-per-page-only view).
• Job Title "CEO" → 71 real matches across 8 pages, correctly including variants like "Founder / CEO", "CEO&Founder", "Co-Founder and CEO", "Ceo" (lowercase), etc.
• Company "Anote" + Job Title "CEO" combined → 6 matches, correctly excluding Anote members with other titles.
• Company matching is case-insensitive against real data (an "ANOTE" row matched the "Anote" filter).
• Clearing an individual chip, "Clear filters", empty state, and pagination-preserves-filters all behaved as expected against the live dataset.

Tests completed

• 32 tests passing across 3 suites (npm test), including 2 new suites: memberFilters.test.js (pure-logic unit tests) and ArmorMembers.test.js (integration tests covering Company/Job Title filtering, AND-combination with search, case-insensitivity, keyword matching, page-reset-on-filter-change, filter persistence across pagination, clear-all and individual-chip clearing, empty state, deduped/blank-free dropdown options, URL hydration/sync, and loading/error states) — using a 25-row fixture specifically constructed so matches span pages 1, 2, and 3, to catch any "only filters the visible page" regression.
• npm run lint: 0 errors (160 pre-existing warnings elsewhere in the repo, none introduced by this change).
• npm run build (CI=false, matching this repo's own deploy.yml/ci.yml convention): succeeds.
• Manually verified in a live browser against the real dataset (see above).

Known limitations

• ArmorMembers.js still calls the Google Sheets API with a hardcoded API key committed in source (pre-existing, unrelated to this PR — same pattern that was previously fixed in ArmorReferrals.js). Not addressed here to keep this PR focused on the filtering feature.
• Because filtering happens synchronously in memory (the whole dataset is already loaded), there's no separate "filtering" loading spinner — only the initial full-dataset fetch shows a loading state. This is a deliberate scope choice, not an oversight.