Tests: frac generator schema test + screenshot-dashboard panel assertions

> **Part of the 4-issue set (Epic → A → B → C → D), one PR.** Spec: [https://github.com/aws-samples/sample-edge-to-cloud-digital-ops-workshop/blob/feat/frac-load-dashboard-timeseries/docs/superpowers/specs/2026-09-09-frac-load-and-dashboard-timeseries-design.md](https://github.com/aws-samples/sample-edge-to-cloud-digital-ops-workshop/blob/feat/frac-load-dashboard-timeseries/docs/superpowers/specs/2026-09-09-frac-load-and-dashboard-timeseries-design.md).

## D — Tests & verification

### Deliverables
1. **Generator schema test**: a small test (or `--dry-run`-based assertion) confirming `frac-msk-load.py` emits `{sensor,value,unit,ts_ms,site_id}` with `site_id` starting `ws-slot` and numeric `ts_ms`.
2. **Dashboard screenshot assertions**: extend `e2e/screenshot-dashboard.mjs` (which already dumps panel text to stdout) to confirm the three new panels (Ingest Volume, Freshness over time, Query Latency over time) render.

### Acceptance (agent, offline)
- The generator schema test passes without network (uses `--dry-run`).
- `e2e/screenshot-dashboard.mjs` references the three new panel titles in its assertions/output parsing.

**Blocked by A and C.** Live screenshot against a real slot stays with the local credentialed session.

Start with frac-msk-load.py and its --dry-run path to verify the emitted schema, then inspect e2e/screenshot-dashboard.mjs, which already parses panel text. Done means the offline generator assertion checks sensor, value, unit, ts_ms, and site_id, and the dashboard assertions reference Ingest Volume, Freshness over time, and Query Latency over time; work is blocked by A and C.

Written by the indexing model from the issue text.