Orchestration Instance Priority and Preemption

[wsugarman]

This is somewhat related to #1803.

For consumption plans and for environments where scale is flexible, all orchestrations and their activities can be treated equally as new workers can be spooled up to process any additional load. However, when resources are more constrained, some orchestrations may be a higher priority than others. For example, there could be some system-critical timer-triggered orchestrations that are stuck behind the activities of something much less important.

While preemption can be built by users (perhaps through the use of IDurableOrchestrationClient and external events), I think it would make sense to provide priority and preemeption in the framework, especially with Azure Durable Functions available in other hosting environments like Kubernetes. Kubernetes itself has a similar abstraction to Durable Functions in Jobs which can be prioritized through the use of PriorityClasses associated with the pods.