WIP: actor based state handling, streaming container stats and podman events

[towe75]

Goals of this branch:

• avoid mutexes for taskhandle storage
• replace per-task container stats polling with a single stats stream
• replace per-task container dead check with a podman event stream observer
• fix a long standing race condition with very short living containers

General idea is to replace the mutex based task store with a message based actor. The actors mailbox, a simple channel, serializes all requests and makes explicit locking unnecessary in a lot of cases.

It's expected that these changes greatly improve the performance of this plugin.

One bug that i found while working on the branch: containers/podman#8864

[hashicorp-cla]

All committers have signed the CLA.

[apollo13]

Hi @towe75, this looks great. What would be needed to move this forward? Also I See that you are restarting the stats/event streams after some delay if the fail. Wouldn't that mean that we loose events?