Re-running a single task triggers the execution of all post_deployment tasks

[Simon Pasquier]

Detailed bug description:

When a user runs this command:

$ fuel node -env <ENV_ID> node --node-id <NODE_ID> --tasks setup_repositories

This isn't only re-executing the setup_repositories task but also all the post_deployment tasks. This is particularly painful for plugins.

Steps to reproduce:

1. Deploy an environment with the example plugin from the fuel-plugins repo.
2. Once the environment is deployed, run this command on all controller nodes:
rm -f /tmp/fuel_plugin_example_v4_sh
3. run the following command for all controller nodes:
$ fuel node -env <ENV_ID> node --node-id <NODE_ID> --tasks setup_repositories
4. Check the presence of /tmp/fuel_plugin_example_v4_sh on the controller nodes.

Expected results:

Only the setup_repositories task (and its required tasks) should be re-executed so /tmp/fuel_plugin_example_v4_sh should be missing.

Actual result:

Post-deployment tasks are re-executed so /tmp/fuel_plugin_example_v4_sh is present.

Reproducibility:

Always

Workaround:

None

Impact:

This breaks the principle of least surprise: if you ask for a specific task to be re-executed, you don't expect that unrelated tasks will be executed too.

This makes the life of plugin developers more complex because they need to make sure that the unattended execution of their post_deployment tasks doesn't break the plugin's functionality. We have a significant track record of bugs related to this behavior:
https://bugs.launchpad.net/lma-toolchain/+bug/1570850
https://bugs.launchpad.net/fuel-plugins/+bug/1529642
https://bugs.launchpad.net/fuel/6.1.x/+bug/1527330

Description of the environment:
 Operation system: ubuntu
 Versions of components: MOS8
 Reference architecture: HA
 Network model: Neutron
 Related projects installed: None
Additional information: None