Move all Batch node initialization and clean-up operations to the runner

[giventocode]

Overview

The BatchScheduler in the TES server generates, uploads, and manages the execution of bash scripts that perform the following operations on the execution node:

• Move the Docker data/image directory to the data disk.
• Clean-up/prune Docker container images.

However, these operations should be performed before the execution of a container command by the TES runner.

AC:

• All operations must be implemented in a new class that executes these operations and encapsulates their details -e.g. ExecutionEnvironmentManager.
• Must move the current network configuration from the DockerExecutor class to the new class.
• Must implement proper observability (logging, without exposing secrets or sensitive information,) of the execution of these operations to facilitate troubleshooting.
• Operations must be idempotent.
• Must integrate the new class with the DockerExecutor as a dependency that is called before the execution of task command on a Docker container.
• Should consider using AutoRemove = true in the host configuration instead of docker pruning commands.

[ngambani]

@giventocode can we link this issue to the TES runner https://github.com/microsoft/ga4gh-tes/issues/148? Or should this be separate from the runner implementation?

[BMurri]

@MattMcL4475 #515 implements Clean-up/prune Docker container images. but does not implement Move the Docker data/image directory to the data disk.

If we are going to retain the current behavior of having TES generate start tasks for the docker configuration part, then we should state that and reclose this issue. Alternately, we could open a new issue linking back to this one to move the docker configuration task from the TES server to the TES runner. Otherwise, the issue is only half completed.