Overview tab CPU chart shows SQL-only CPU while alert fires on combined CPU

[erikdarlingdata]

Summary

The Overview tab's "% CPU" chart plots only sqlserver_cpu_utilization, but CPU alerts evaluate combined CPU (sqlserver_cpu_utilization + other_process_cpu_utilization). This causes user-facing confusion when an alert fires at, say, 95% but the Overview chart never shows above ~60%.

The "CPU Utilization" sub-tab correctly stacks both series, so the data is visible there — just not on the landing/Overview chart users glance at first.

Reported by a user as erikdarlingdata/PerformanceStudio#349 (filed against the wrong repo — actual code lives here).

Repro

1. Run a workload where SQL Server CPU is ~60% and other processes consume ~30%.
2. Set CPU alert threshold to 80%.
3. Alert fires reporting ~90% — correct, that's combined.
4. Open Overview tab → "% CPU" chart never exceeds ~60% → looks like the alert is bogus.
5. Open "CPU Utilization" sub-tab → stacked chart shows the true combined total → matches the alert.

Code pointers (from exploration)

• Alert evaluation uses combined CPU — Lite/Analysis/DuckDbFactCollector.cs ~lines 414–472 aggregates both series; Lite/Services/LocalDataService.Overview.cs ~line 190 colors CPU red on the alert metric.
• Overview chart plots SQL-only — Dashboard/Views/ServerTab.Charts.cs ~line 475:

  dataList.Select(d => (double)d.SqlServerCpu)

  The query in Dashboard/Services/DatabaseService.ResourceMetrics.cs ~lines 230–289 already retrieves total_cpu_utilization, so the data is available — the binding just isn't using it.
• Daily summary "High CPU Events" in Dashboard/Services/DatabaseService.Overview.cs ~lines 108–112 and ~136 also use sqlserver_cpu_utilization >= 80/90 thresholds (SQL-only). These should probably switch to total_cpu_utilization for consistency with the alert too.

Proposed fix

1. Overview chart: plot total CPU (or stack both series like the CPU Utilization sub-tab does — Erik's call which presentation is better).
2. High-CPU event counters in DailySummary: switch to total_cpu_utilization so they match what triggered the alerts.
3. Verify no other UI uses SQL-only CPU in a way that implies "system CPU."

Once that ships, PerformanceStudio#349 can be closed.