Timeout requests if workers never spawn

[rauchg]

https://github.com/LearnBoost/up/blob/master/lib/up.js#L210

In the rare case where all workers failed (and also fail to respawn), prevent memory bloating of the distribute process.
But also make sure first node-http-proxy can't do this for us.