feature: Provide playbooks with better control over restarts

[glennpratt]

Summary

Currently, this uses hacks to perform rolling restarts within the role. There is no ability to configure the loop + include hack, it spams the logs and wastes resources (at least according to other reports). Most importantly, it's just painfully slow on large clusters with multiple pools.

If this role was also shipped as a collection, we could add playbooks that could make use of the serial, throttle and hosts ordering features of plays to have fine grained control over restarts.

Ideally, invoking the role through these playbooks could expose those fields through vars and skip any role loop based restarts.

This can be created completely external to this repo, but changes to this role would be required to disable loops that would then be implemented in plays.

I plan to work on a proof of concept soon; just opening this to start the discussion early.

Issue Type

Feature Idea