TODO

- TwitterIdFetcher.getIds() should be called asynchronously.
  Currently the stream processing is interrupted and doesn't resume
  until we've fetched the next set of twitter ids.  The ids should be
  fetched asynchronously and handed off to the processing threads.

- TwitterIdFetcher.getIds() may have a partial failure, i.e., it may
  only successfully fetch ids from a subset of sources.  In that case,
  we should proceeed to follow those ids, but should retry the
  complete fetch periodically instead of waiting for the end of the
  processing interval.

- Add support for setting the "delimited" parameter.

- Tests.