feat(grafana): add fleet health dashboard

[mislayming]

Summary

• add a provisioned and embedded Ongrid Fleet Health (Troubleshooting) Grafana dashboard
• cover device offline counts, top last_seen age, edge connections, scrape failures, manager runtime pressure, Prometheus series cardinality, prom remote_write calls, and alert evaluator ticks
• keep the dashboard portable by using only generic ongrid/Prometheus metrics and no environment-specific hostnames, URLs, or screenshots

Dependency

• Draft until fix(alert): recover device offline detection #164 is merged, because this dashboard uses the stable device_last_seen_timestamp_seconds{device_id} metric added there.

Tests

• jq empty internal/manager/biz/grafana/dashboards/fleet-health.json deploy/install/grafana/provisioning/dashboards/json/fleet-health.json deploy/grafana/provisioning/dashboards/json/fleet-health.json
• go test ./internal/manager/biz/grafana ./internal/pkg/grafana

Author confirmation

• I have read and agree to the project contribution guide in CONTRIBUTING.md.

[singchia]

#165 这个方向可以。新增一个独立的 Ongrid Fleet Health Grafana dashboard 没问题，它和现有 ongrid-monitor 不是替代关系：

• ongrid-monitor：跟 Ongrid Monitor 页保持一致，主要看主机资源趋势和用户自定义面板；
• fleet-health：偏排障入口，集中看设备离线、last_seen age、edge connection、scrape failure、remote_write、alert evaluator、manager/Prometheus 压力。

这样作为 troubleshooting dashboard 是有价值的。

但这个 PR 现在不能直接进，因为它同时带了 #164 的内容。请先等 #164 合并后 rebase，或者重新拆一个干净分支，只保留 dashboard 相关文件：

• deploy/grafana/provisioning/dashboards/json/fleet-health.json
• deploy/install/grafana/provisioning/dashboards/json/fleet-health.json
• internal/manager/biz/grafana/dashboards/fleet-health.json

其它 alert pipeline、device offline metric、migration、rule preset 变更应该都属于 #164，不要混在这个 PR 里。

如果要把体验再做完整一点，后续可以单独加一个入口：Monitor 页继续打开 ongrid-monitor，旁边再给一个 Fleet Health / troubleshooting 链接打开 ongrid-fleet-health。这个可以另开 PR，不阻塞本 PR。

[singchia]

Reviewed. I would not merge this PR in its current form.

The Grafana fleet dashboard itself looks fine as an additive panel, but this branch also includes the alert/device-offline changes from #164. Please split or rebase this PR so it contains only the dashboard files. We should decide/fix #164 separately before taking these dashboard changes.