Fix PARALLEL executorGroup to ensure sequential execution for single DiscoveryRequestStreamObserver

[MarcinFalkowski]

Currently we use one ThreadPoolExecutor shared for all DiscoveryRequestStreamObservers as an ExecutorGroup: https://github.com/allegro/envoy-control/blob/master/envoy-control-core/src/main/kotlin/pl/allegro/tech/servicemesh/envoycontrol/ControlPlane.kt#L105
(only when corresponding property is set to PARALLEL, which is not the default)

This approach is not valid because it will lead to sending XDS responses out-of-order for given DiscoveryRequestStreamObserver.

We should switch our parallel ExecutorGroup implementation to multiple, single-threaded ThreadPoolExecutors. This way we will ensure that single DiscoveryRequestStreamObserver is running sequentially, but many DiscoveryRequestStreamObservers may run in parallel.

Implementation of such ExecutorGroup may look like this (not tested):

 class SequentialExecutorGroup(
    threads: Int,
    singleThreadedExecutorFactory: (Int) -> ExecutorService
) : ExecutorGroup {

    private val executors = (0 until threads).map(singleThreadedExecutorFactory)
    private val counter = AtomicInteger(0)

    override fun next(): Executor {

        val index = counter.getAndUpdate { c ->
            val next = c+1
            if (next >= executors.size) {
                0
            } else {
                next
            }
        }

        return executors[index]
    }
}