Add failhard option to salt orchestration

[karanjad]

Feature request;
Say in a long orchestration and one component failed and due to the requirements of the environment, it was necessary to stop the entire orchestration at the point of that failure.

[basepi]

I think orchestrate should already support failhard for specific pieces of the orchestrate run (much like state runs do, failhard is a global state argument). But we should definitely also have a global failhard option for orchestrate if it doesn't already exist.

[karanjad]

Lets say you have a scenario where you do not want to set a global failhard and the orchestration is taking place across different minions. In this situation, a failhard from one minion fails to stop the orchestration.

[cachedout]

This was actually already present but just needed a small fix to the state compiler to make it functional. I've made the necessary change in #23389. After that is merged, failhard will work in orchestrations.