For clusters, allow --wait to be passed down

[emschwar]

This comprises two related changes:

• For individual servers, use --wait to timeout a start or stop effort.
• For clusters, pass the --wait parameter down to each cluster's stop/start/restart
  command.

(Fixes #310)

[emschwar]

Running specs shows this failure:

1)
Errno::EPROTOTYPE in 'Thin::Server robustness should not crash when header too large'
Protocol wrong type for socket
spec/server/robustness_spec.rb:19:in `write'
spec/server/robustness_spec.rb:19:in `block (4 levels) in <top (required)>'
spec/server/robustness_spec.rb:18:in `times'
spec/server/robustness_spec.rb:18:in `block (3 levels) in <top (required)>'
spec/server/robustness_spec.rb:12:in `times'
spec/server/robustness_spec.rb:12:in `block (2 levels) in <top (required)>'

I think it may be just my environment, since I'm pretty sure I didn't change anything that would have affected this test, but I'm happy to make any additional fixes if it's a real problem.

[ioquatix]

Can you please rebase this on master?

[emschwar]

Done.

[ioquatix]

Errno::EPROTOTYPE in 'Thin::Server robustness should not crash when header too large'

This is actually a bug in the MacOS network stack due to deferred resource allocation - you get a very weird error because that is actually the point where the socket is created, and the error occurs... it comes out of an "unrelated" system call.

Thanks for rebasing.