Orca lead - Add capability to specify how long a test will run

[lancelakey]

There's too much variance in the results from short tests. I need to be able to somehow make the tests longer or somehow control how long they are.

This is related to my rant in #17

[dyoder]

I've said this already in discussions, but I figured I'd note it for the record: Orca is intended more to be able to do simulations, rather than simply blasting the API. The "right" way to do longer tests is to make the simulations more long-lived (or, perhaps more accurately, to enhance Orca to support that better).

That said, adding this one feature is probably okay. I just don't want to turn this into a crappy version of httperf. Otherwise, obviously, we'd just use httperf.

It might sort of seem funny right now, because our tests aren't very interesting. (So maybe we should, in fact, just use httperf, running across different nodes.) But it starts getting more interesting quickly once we start saying things like, "okay, let's see what happens when someone starts logging a bunch of bogus events" and we add that to our simulation.