Added bootup-err checking logic.

If the job uses more than 1 device (shards >= 2), the process pool
creates child processes for parallelized flashing. So the main "job"
process needs to check if any of its devices went to bootup-err state
after the job is done.

Test: > device --lease=True
Bug: 77997275
Change-Id: Id9ae397fd4d2fc281a798eb6b868a5cf99bd533f