proposal - allow planning node changes

In an effort to maximize resource utilization in my nomad clusters while maintaining confidence that certain jobs can be placed, it would be a great benefit if I would have a tool for running plan on removing a node(s) from the cluster just like I have for placing a job or changing one. This tool would report the plan/placements that would happen if the specified set of nodes would go down/would be drained.

With this in place I could manage the provisioning/sizing of my different node classes (with a periodic job for example) provisioning/decommissioning capacity in a manner that maximizes utilization but still retains enough capacity to prevent placement failures on class/constraint basis for those jobs that are most important for me.

Start by reading the existing planning behavior for placing or changing jobs, then trace how node drain or failure affects allocations. Define the planning input for a specified set of nodes and verify that the result reports the placements that would occur after those nodes go down or are drained.

Written by the indexing model from the issue text.