Fix replication Javascript test

[nickva]

The test was flaky for a variety of reasons:

• waitForSeq only waited for 3 seconds and on failure it never explitly
  indicated an error and just waited for the comparison below to fail. So made
  it wait for 30 seconds and also throw an exception right away if it fails.

• Last waitForSeq was used after task was canceled. So it just wasted time
  waiting until timeout as the task was null. So created a function to
  wait for task to be null.

• waitForSeq spun in a tight do/while loop querying _active_tasks. In some test
  environment with minimal CPU resources that's not the greatest thing to do.
  So made it wait for 0.5 seconds between retries.

• waitForSeq waited for replication task's through_seq value to match source
  update sequence from source db info. Those don't necessarily match. Instead
  made waitForSeq use the changes feed last sequence since that's what the
  replication task uses to update through_seq.

[nickva]

Tested on a CPU and disk resource constrained VM

CPU resources slowed down to 9% and disk throughput limit set at 100KB/sec

VBoxManage bandwidthctl ${VM} set Limit --limit 100KB

Seems to pass when before it failed:

 test/javascript/run replication.js
test/javascript/tests/replication.js   pass
=======================================================
JavaScript tests complete.
  Failed: 0.  Skipped or passed: 1.

[davisp]

+1

[wohali]

Sadly we have a new failure scenario now. I'm going to reopen #634.