Add initial delay and scale wait parameters with ntasks

[dmargala]

This PR makes a few adjustments to help with shared run startup at "high" concurrency per node.

The wait_timeout and wait_poll_interval parameters are log2 scaled with ntasks (note this is "number of tasks per node").

For example, with wait_timeout=10 and wait_poll_interval=0.2:

ntasks	wait_timeout	wait_poll_interval
1	10	0.2
2	20	0.4
4	30	0.6
8	40	0.8
16	50	1.0
32	60	1.2
64	70	1.4
128	80	1.6
256	90	1.8

Also, the wait poll loop is converted to a "do-loop" style which includes an initial wait before checking for the existence of the container to give the container a chance to start up before polling for it's existence.