Set multinodetests to run on 'all' builds

[Aaronontheweb]

Minor change, but going forward all multinodetests will run on build.

This is a big step towards getting 1.1 out the door.

[Aaronontheweb]

Builds ran successfully several times on local machine, but experienced a hang here:

[00:53:26] :     [Step 1/1] [NODE5][6/23/2015 12:53:26 AM][DEBUG][[akka://MultiNodeClusterSpec/user/TestConductorClient]]: transition AwaitDone, Akka.Remote.TestKit.ClientFSM+Data -> Failed, Akka.Remote.TestKit.ClientFSM+Data
[00:53:26] :     [Step 1/1] [NODE5][6/23/2015 12:53:26 AM][DEBUG][[akka://MultiNodeClusterSpec/user/TestConductorClient]]: processing Event: <<Terminated>: [akka://MultiNodeClusterSpec/user/$b] - ExistenceConfirmed
[00:53:26] :     [Step 1/1] [NODE5][6/23/2015 12:53:26 AM][DEBUG][[akka://MultiNodeClusterSpec/user/TestConductorClient]]: unhandled event <Terminated>: [akka://MultiNodeClusterSpec/user/$b] - ExistenceConfirmed
[00:53:26] :     [Step 1/1] [NODE5][6/23/2015 12:53:26 AM][DEBUG][[akka://MultiNodeClusterSpec/user/TestConductorClient]]: transition Failed, Akka.Remote.TestKit.ClientFSM+Data -> Failed, Akka.Remote.TestKit.ClientFSM+Data

[Aaronontheweb]

I upgraded from a 2-core, ~4GB machine to a 4-core, ~8GB machine and that appears to have made the difference for the TestConductor, which would just deadlock at the very beginning of the multi-node tests before. I think it requires a minimum level of concurrency that it couldn't achieve with just 2 cores before.

The good news is that the new machines are still low-powered enough to expose race conditions in the tests (hence #1089,) so I think this is a happy medium.