Docker container metadata fetching is blocking, which can result in dropped system calls

[mstemm]

Currently, when sysdig notices that a docker container has been created, it does a blocking fetch of metadata from the docker socket. This can take up to hundreds of milliseconds for the docker daemon to respond with the container's metadata, and on busy systems this can result in dropped system calls.

We should make fetches of container metadata asynchronous and have a separate "container event" that notes that a container has been created. We already have an internal CONTAINER_JSON event that is used for trace files, we just need to make this event visible.