Fix Lite perfmon chart bugs and Dashboard ScottPlot crash handling

[erikdarlingdata]

Summary

• [QUESTION] Lite - Perfmon graph spikes & multiple values at same timestamp #544: Lite perfmon charts showed jagged lines and spikes because instance_name wasn't aggregated and gap detection was missing from the delta calculator
• [BUG] #545: Dashboard crashed with XamlParseException when SkiaSharp native libs couldn't load — now shows a helpful error message instead

Changes

• Lite chart query: SUM(delta_cntr_value) GROUP BY collection_time so multi-instance counters (e.g. "Query optimizations/sec" with "default"/"internal") render as a single clean line
• DeltaCalculator: Added timestamp tracking and gap detection — if >5min since last cached value, returns 0 delta instead of an inflated value spanning the entire gap. All seed methods now store timestamps from DuckDB.
• Perfmon collector: Passes collectionTime and maxGapSeconds: 300 to delta calculator. Fixed sample_interval_seconds from hardcoded 600 to 60.
• Dashboard: Wrapped ServerTab construction in try-catch with error message pointing to VC++ Redistributable download

Test plan

• Launch Lite, connect to a server, let it collect for a few minutes
• Check Perfmon tab — counters like "Query optimizations/sec" and "Network IO waits" should show single smooth lines
• Stop and restart Lite — first data point after restart should not spike
• Dashboard opens server tabs normally (no regression from try-catch)

Closes #544, closes #545

🤖 Generated with Claude Code