Dashboard filters for drift and better failure semantics

Allow filtering out drift detection runs from dashboard metrics and improve failure handling.

We have drift detection enabled on almost all stacks, close to 1000, scheduled at midnight. These runs queue and heavily skew dashboard metrics.
• Average and median run duration charts show values around three hours, dominated by queued drift runs, so they no longer reflect real interactive runs.
• The dashboard also shows hundreds of drift runs as failures, which does not match the real state, so the failure metric is not reliable for us

Would it be possible to filter this in the dashboard?