Revamp number of iterations vs. number of configs tracking

[bpkroth]

Discussed offline:

• previously (Add basic support for repeating a trial config. #642) we artificially increased the iteration count so that when we repeated trials it wouldn't lessen the number of configs we sampled
• now that we're separating the optimizer (suggest config) and scheduler (repeat of configs), the we can
  • move the tracking of number of configs suggested to the suggest() routine in the Optimizer (Increment iteration counter in .suggest() instead of .register() #711), and
  • number of trials executed to the Scheduler
• that opens up the door for more flexible stopping conditions (Implement more flexible "Stopping Conditions" #427)
• to do that, we should also cleanup our terminology and
  • rename max_iterations to num_config_suggestions, and
  • add a new max_trials argument (separate PR though I think)

Originally posted by @bpkroth in #710 (comment)

[motus]

I think we've fixed this one in #713 and #719