Add ability to disable certain Heat Deployments to a set of servers

The ability to not rerun all SoftwareDepoyments on every stack-update is commonly requested. The most common use case for this request is scaling up/down compute nodes. In that scenario, it's not strictly necessary to run 5 passes of puppet on every other role in your deployment just to add a single compute node. You should be able to add a compute node to your deployment and leave controllers, etc completely untouched.

A generic way to specify what Deployments get executed on what servers on stack-update would be a useful addition to the framework.

Although, we'd have to document the feature sufficiently about how it is meant to be used as it would be possible to abuse the feature such that stack-updates fail in strange ways. We'd have to explicitly document how to use it for specific use cases (such as scale up of compute nodes), and caution against other uses where users could potentially do unintentional harm.