Refactor batcher and fix race condition

[eapache]

Fixes #46

- Run CI tests with race detection enabled so CI will catch issues like this in the future
- Replace done channel with a WaitGroup to fix a race condition and to
  properly wait for *all* active batches during shutdown
- Use time.AfterFunc() instead of manually spawning a goroutine and
  calling sleep()
- Improve a few method names and doc comments
- move some logic from batch() to submitWork() so it's always obvious
  what is or isn't guarded by the lock