feat: wire TUI dashboard panels to live data (closes #134)

[luceinaltis]

Summary

Closes #134. Makes the Textual TUI dashboard functional by wiring panels to the live data layer instead of showing only static config snapshots.

• OverviewPanel: fetches live prices for portfolio holdings and the watchlist via DataRegistry.async_get_with_fallback(PriceProvider, "get_price", …); per-holding P&L computed when prices are available. Replaces the static "S&P 500 / NASDAQ / VIX" placeholder with a live Watchlist mini-table.
• PortfolioPanel: uses RiskCalculator.build_snapshot() to render market values, weights, and unrealized P&L once all holdings have prices; falls back to a cost-basis view for partial data. A total-value label shows either live portfolio value or cost basis.
• WatchlistPanel: fetches live prices for each ticker on refresh.
• AlertsPanel: replaces the static placeholder with a DataTable backed by AlertStore, with hot-reload via mtime checks (picks up alerts created by qracer repl without restarting the dashboard).
• TasksPanel (new): reads TaskStore.get_active() and renders schedule + human-readable next run time (in 5m / in 2h / due). Added to the DASH sidebar group with hotkey 5.
• QracerDashboard accepts optional data_registry, alert_store, task_store kwargs and injects them into panels both on compose and on sidebar navigation.
• qracer dashboard CLI now calls _build_registries() and passes ~/.qracer/alerts.json + ~/.qracer/tasks.json stores to the app.

Panels keep their placeholder behaviour when no registry/store is provided (useful for tests and users without live data) — every cell that depends on live data falls back to —.

Test plan

• uv run pytest tests/dashboard/ → 49 passed (14 new tests: refresh_data behaviour for Overview/Portfolio/Watchlist with a _FakeRegistry, partial-price fallback, AlertsPanel hot-reload, TasksPanel empty/populated/hot-reload, _format_next_run edge cases)
• uv run pytest tests/cli/ tests/test_cli.py tests/test_alerts.py tests/test_tasks.py tests/test_watchlist.py tests/storage/ tests/risk/ → 238 passed total
• uv run ruff check qracer/dashboard/ qracer/cli.py tests/dashboard/ → clean
• uv run pyright qracer/dashboard/ qracer/cli.py → 0 errors
• Manual smoke test: qracer dashboard with a populated portfolio and watchlist — verify prices appear within one refresh interval and switching panels keeps live data.

Notes

• Refresh cadence is unchanged (REFRESH_INTERVAL_SECONDS = 5.0).
• Async price fetches go through DataRegistry.async_get_with_fallback, matching the pattern used by AlertMonitor, so provider priority and fallback are consistent across the app.
• Individual ticker fetch failures are logged at DEBUG level and skipped rather than blanking the whole table.