Make reconcile error panels honor Grafana time range

[swills]

This PR updates the cluster dashboard reconcile error panels so they honor the Grafana-selected time range instead of using raw cumulative counters.

This closes #62 and is related to #48

The dashboard currently uses raw controller_runtime_reconcile_total{result="error"} counters for several reconcile error panels. Examples:

clamp_max(max(controller_runtime_reconcile_total{namespace=~"$operatorNamespace", result="error", controller="backup"}), 1)

and:

max(controller_runtime_reconcile_total{namespace=~"$operatorNamespace", result="error", controller="backup"})

Since controller_runtime_reconcile_total is a monotonically increasing counter, these queries effectively answer:
"Has this controller ever seen a reconcile error since the metrics process started?" This makes the panels stay red/noisy long after transient issues have cleared.

This PR wraps those counters with increase(...[$__range]). For example:

clamp_max(max(increase(controller_runtime_reconcile_total{namespace=~"$operatorNamespace", result="error", controller="backup"}[$__range])), 1)

and:

max(increase(controller_runtime_reconcile_total{namespace=~"$operatorNamespace", result="error", controller="backup"}[$__range]))

Using increase(...[$__range]) makes the panel reflect reconcile errors that occurred within the currently selected Grafana time range. That means the panel semantics become:

"Did any reconcile errors occur during the selected dashboard time range?"

instead of:

"Did any reconcile errors ever occur since the operator process started?"

This works well for interactive dashboard use because it naturally follows the Grafana time picker:

• 5m -> last 5 minutes
• 1h -> last hour
• 24h -> last day

This PR updates:

• the top overview reconcile error panel queries
• the related reconcile error stat/timeseries panel queries elsewhere in the dashboard