Add a command to recover from an inconsistent state

[PaulFarault]

In some edge situations (due failure with Ansible, machine crash, etc.), the database may present an inconsistent state where the last deployment state stays RUNNING while no Ansible task are running. However, only deployment with a FAILURE state can be resumed using the tdp plan resume command.

We should add an option/command/sub-command to recover from this state. It should allow to:

• Update the last deployment state from RUNNING to FAILURE
• Update the operations:
  • If an operation is RUNNING, update it to FAILURE and set the following operations to HELD
  • If an operation is PENDING or HELD, update it to HELD and set the following operations to HELD