Split provisioning and execution

[ispasov]

Description

• Split provisioning and execution

Previously the provisioning and execution were bundled together in the same func.
This leads to several issues:

1. When the execution fails, the provisioning tried to deploy a new VM, thinking the provisioning of the VM failed.
2. When the execution fails due to ssh connection, a new VM was either deployed or the running one was deleted, although it was running an actual active job.

Now the provisioning retries until a VM and a runner are present.
Then the execution is performed.
The resources are cleaned if the execution completes. If there is an SSH error then the VM is not deleted and it is relied on the JobCompleted event to clean up the resources.
This ensures active VMs are not deleted and the jobs are not lost.

Finally, add more logging

Testing

Case 1

1. Start the runner make run
2. Run a job and ensure it completes

Case 2

1. Ensure you have a job that takes at least 10-15 seconds.
2. Start it
3. Follow the runner logs, the moment you see that the runner has started, stop the VPN connection.
4. The job should complete and the runner should be deleted together with the VM

Case 3

1. Use the same job
2. The moment the runner starts the job, delete the VM
3. You should not see any attempts to cleanup the VM
4. Stop the runner and cancel the job

Case 4

1. Use a job that fails
2. Ensure the runner correctly deletes the VM and the job fails successfully

[relwell]

It might be good to wrap this into a private function so that the behavior is a little more semantic