These changes allow for a Node to check configuration values and
potentially veto bringing up a Node that is otherwise in demand.  For
example the Lab Manager plugin could be extended to allow for only 5
slaves in a given cloud to be online at given time.

Signed-off-by: Tom Rini <tom_rini@mentor.com>

- AbstractProject.getBuildingUpstream/getBuildingDownstream()
  should be symmetric.
- It's unlikely for us to add another category of items,
  so reverted getUnblockedItems() implementation to the older one
  that was more efficient.
- ItemList was meant to be internal, so keep it as such,
  at least for now.

Changed logic to build list of unblocked items in the queue safely so it will not break if a new category of queue items is added

[FIXED JENKINS-8929] Changed the definition of a building downstream project to one that is Building, Waiting, Pending or Buildable
but NOT blocked.  This prevents the deadlock condition where upstream project A is blocked on downstream project B, which is
blocked on upstream project A - now in this scenario B will remain blocked but A will build, eventually freeing up B.

items, which upgrades the versions used for maven-hpi-plugin,
maven-stapler-plugin and maven-localizer-plugin

upgrade to maven 3.0.3 (will prevent issue when pom parsing phase : http://jira.codehaus.org/browse/MNG-4963)

take care about using settings from the remote node

If you had two clouds and only the second one could provision a label, it'd never get that far - check to see if a cloud can provision a label before trying to do so.

I wasn't initially planning to do this, but it turns out more people needed this feature than I thought.

I left the text as is, but defined a mechanism to override this in the plugin, in a possibly context sensitive way. For example, a plugin can replace this text on a job-by-job basis, or by looking at some property/builder/etc.