Hub agent uses only 1/10 of workers

[ahmetb]

We ran into an incident today due to Kubefleet hub agent failing to drain a large workqueue.

We bumped our worker count from 40 → 80 but we were still seeing extremely slow recovery ( O(hours) ).

It turns out Hub agent actually uses 1/10 of workers specified:

kubefleet/cmd/hubagent/workload/setup.go

Line 518 in 39ed26b

ConcurrentPlacementWorker: int(math.Ceil(float64(opts.MaxConcurrentClusterPlacement) / 10)),

This is a big no-no in controller development. Controllers should respect the user input and not mess with it as it breaks the expectation between the user and the system.

cc: @mikehelmick @ArchanaAnand0212