Experiment / Benchmark using "worker" goroutines to run invocations

Right now the worker/actor invocation always run in the calling Goroutine. This can lead to a huge amount of time being spent in runtime.NewStack for each invocation. We should experiment with running a pool of worker routines in `activations.go` and sending them work over a channel to see if that improves performance at all. I suspect it will have a dramatic impact on latency for some workloads.