Adds support for starting multiple managers per host

[ddanielr]

Updates the cluster.yaml format to support setting the number of manager processes per host.

Closes: #6212

If managers should be organized under a resource group then this PR can be easily modified to support that.

[dlmarion]

I don't think we should put managers into resource groups until we define what that means.

As a user I would assume that if a manager was in the test resource group, then it would manage all of the other resources and tables that were also in the test resource group. That might seem wasteful resource-wise if nothing is really happening in the test resource group.

A user might want to then have the manager manage the resources in multiple resource groups. So instead of putting the manager into a resource group, the manager would have resource groups assigned to it to manage.

Right now we have a default resource group that everything is lumped into. Because the root and metadata tables are in the default resource group it has a minimum requirement of 1 tablet server and 1 compactor. If we are going to keep the current multiple manager design, equal distribution of work across managers, it might make sense to have a system resource group that the managers, gc and monitor use (and cannot be changed) with the 1 tablet server and 1 compactor minimum.