agent: don't treat downscale denies as failed request (#927)

Fixes #926, a follow-up for the #770.

The defintion for the VM stuckness was changed to include denied
downscale request in the following commit:

    commit fdf0133
    Author: Shayan Hosseini <[email protected]>
    Date:   Sat Apr 6 09:25:01 2024 -0400

    agent: track more liveness in vm-stuck metrics (#855)

This resulted in the consistent firing of the alert.

We should actually treat the denied downscale as part of the normal
operation. This can happen due to mismatching policy of what is an
acceptable level memory usage in autoscaler-agent vs vm_monitor.

Signed-off-by: Oleg Vasilev <[email protected]>

pkg/agent/execbridge.go

@@ -169,8 +169,10 @@ func (h *execMonitorHandle) Downscale(
 
 	result, err := doMonitorDownscale(ctx, logger, h.monitor.dispatcher, target)
 
-	if err == nil && result.Ok {
-		h.runner.recordResourceChange(current, target, h.runner.global.metrics.monitorApprovedChange)
+	if err == nil {
+		if result.Ok {
+			h.runner.recordResourceChange(current, target, h.runner.global.metrics.monitorApprovedChange)
+		}
 	} else {
 		h.runner.status.update(h.runner.global, func(ps podStatus) podStatus {
 			ps.failedMonitorRequestCounter.Inc()