feat(ui): wire Export/Import job cards to tenant-scoped counts (#554)

[andyguz17]

Closes #554.

What

• The Export jobs card now shows the tenant's real running / queued bulk-export counts (it was a hardcoded "13 / 1 queued").
• New Import jobs card with the tenant's active bulk-submit count, linking to /ui/bulk-import.
• Both cards render an explicit "unavailable" state (em dash + aria-label) when the running backend has no job store or the count cannot be read — never a fabricated zero.
• Both cards are links (<a class="card stat stat--link">): Export → /ui/bulk-export, Import → /ui/bulk-import.
• Stat grid gets a stat-grid--5 modifier (5 cols on desktop, 3 ≤1100px, existing 2-col rule ≤900px).

How

• persistence: BulkExportStorage::count_exports_by_status(tenant, status) (SQLite + Postgres, additive). count_active_exports is untouched — it still backs the concurrency cap.
• observability: DashboardSnapshot gains export_jobs: Option<ExportJobCounts { running, queued }> and import_jobs_active: Option<u64>; None means unavailable.
• rest: StorageDashboardProvider::with_job_stores(..) receives the optional bulk-export / bulk-submit job stores already built in build_app and fills the two fields. running = in-progress, queued = accepted. A read error logs a warning and yields None; no store wired yields None silently.
• ui: DashboardMetrics carries Options, template + CSS + i18n (en/es/de) + HTTP test; e2e page object and spec cover both cards (stat-card count 4 → 5).

Decisions (as requested by the issue)

• Per-tenant job counts come from the authenticated dashboard snapshot, not /metrics; nothing per-tenant is added to /metrics.
• Process-wide export/import Prometheus metrics: not added here — recorded decision; happy to file a follow-up issue if wanted (would instrument the workers, no tenant label).
• The running/queued split is backed by real data (in-progress / accepted); the card copy is unchanged.
• Uptime is untouched (UI Home Page: wire the Uptime stat to Prometheus metrics (implement the metric if missing) #540, @angela-helios); only the placeholder comment next to it was reworded to cover uptime alone.
• 5-column grid kept after a visual check at 1440/1200/1000/800px (dark + light); the es labels wrap to two lines around 1200px but stay within the card.

Coordination

• Depends on feat(ui): Bulk Export workspace — scoped kick-off, filters, active exports #538 for /ui/bulk-export (Export card link) — merge this after feat(ui): Bulk Export workspace — scoped kick-off, filters, active exports #538. If feat(ui): Bulk Export workspace — scoped kick-off, filters, active exports #538 changes the route name, the href in crates/ui/templates/pages/index.html and the e2e assertion are the only two places to touch.
• The issue mentions a max-width: 1081px breakpoint; app.css has no such rule — the existing 900px rule plus a new 1100px rule for the 5-col grid are used instead.

Testing

• cargo fmt --all -- --check, workspace cargo clippy --all-targets --all-features -D warnings (CI allow-list), cargo test -p helios-persistence bulk_export (SQLite unit + Postgres via testcontainers), cargo test -p helios-observability -p helios-rest -p helios-ui, cargo build -p helios-hfs --features ui.
• Manual: hfs on SQLite → /ui shows Export "0 / running (0 queued)" and Import "0 / active".
• Playwright: spec compiles and is discovered (--list); the full browser run relies on CI (ui-tests.yml) — Chromium could not be downloaded on the dev box.

[codecov]

Codecov Report

❌ Patch coverage is 98.75000% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...s/persistence/src/backends/postgres/bulk_export.rs	66.66%	1 Missing ⚠️
...tes/persistence/src/backends/sqlite/bulk_export.rs	98.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!