Disconnected test execution

[richtja]

Is your feature request related to a problem? Please describe.

One of the key features of nrunner architecture should be disconnected test execution. But how the #5794 this feature is still not supported. Right now, the tasks need to communicate with avocado's status server all the time during the task execution. The goal of this feature is to start a task in the spawner's environment and then disconnect from it, and when the task is finished gather logs and results.

This feature will be useful when connection to the spawner's environments is not stable or too expensive to keep it during the whole execution.

Describe the solution you'd like
Let's introduce new Task config variable local_log_dir. This variable will be set by spawner before the Task is executed. When this variable is set, the Task will store any runtime messages to this file to preserve all task runtime information inside the spawner environment to be retrieved by the spawner any time when it is needed.