feat(cloud): push per-miner state alongside snapshots

[proofofprints]

Summary

Wire the existing client::push_miners into the cloud sync loop so the portal's miner_states table gets populated. Until now, the desktop pushed farm-rollup snapshots but never pushed per-miner state, so the web Dashboard's "Total Miners" tile always read 0 even when miners were online.

This is a Phase 2 vertical-slice: capture local miner data → stage on CloudState → push from sync loop → queue/drain on transient failure. Full Phase 2 scope (alert sync, command exec, etc.) is still pending — intentionally not in this PR.

What changed

• src-tauri/src/cloud/mod.rs — Added latest_miners: Mutex<Option<serde_json::Value>> to CloudState, initialized to None. Direct mirror of the existing latest_snapshot field.
• src-tauri/src/poller.rs — build_and_persist_snapshot now also calls a new stage_miners_for_cloud(...) after persisting the rollup. That helper takes the same ASIC + mobile + PoPMiner inputs the rollup already uses, maps each into { minerType, minerId, label, coin, online, hashrate, state }, and writes the result to CloudState.latest_miners. The poller's snapshot_loop now takes PopMinerDevicesState (previously discarded as _popminer_state).
• src-tauri/src/cloud/sync.rs — Added a second push step right after the snapshot push: take() from latest_miners, call push_miners, on transient failure enqueue with kind "miners", on 401/403 set AuthRequired. Extended the queue drain match to handle "miners" items.
• src-tauri/src/cloud/queue.rs — Relaxed the CHECK constraint to include 'miners', plus a one-shot migration for existing installs (SQLite can't ALTER a CHECK, so we rebuild the table when the existing definition is missing 'miners').

Local capture path

The local data source was already there: poller::build_and_persist_snapshot collects asic (Vec<MinerInfo>) from cache.asic_miners, mobile_miners from mobile_state.miners, and now popminer_devices from popminer_state.saved. No new polling logic; we just serialize what the rollup already iterates over into the /ingest/miners shape.

Minor type mapping

• ASIC → minerType: "asic", minerId: ip, label from saved miner or hostname, coin from saved miner.
• Mobile → minerType: "mobile", minerId: device_id, hashrate stays in H/s (no conversion).
• PoPMiner → minerType: "popminer", minerId: mac, coin omitted (no per-device coin on the ESP32 side yet).

The full miner record (model, firmware, temps, etc.) goes into the optional state field for portal display.

Build verification

cd src-tauri && cargo check passes cleanly (no new warnings; pre-existing dead-code warnings unchanged).

Test plan

• Build locally (pnpm tauri dev or release build), log in, add at least one ASIC/mobile/PoPMiner miner.
• Watch the desktop logs for Cloud: miners staged for sync (N entries) followed by Cloud sync: miner state pushed successfully on the next 60s cycle.
• Verify the portal Dashboard's "Total Miners" tile reflects the actual count within one cycle (~60s).
• Pull the network cable / block cloud-api.proofofprints.com briefly and confirm a miners row appears in the local queue (cloud_queue.db); restore connectivity and confirm the queue drains.