Using stamp could possibly cause issues due to UTC datetimes not being
aware of leap seconds.

corrupt session groups can have packet files with multiple APIDs
we handel this by filtering these out and skipping to the next stamp

Because merge was only considering stamp/apid packets for the same
stamp/apid with different byte size were being treated as equal. It now
always treats packets/group with more bytes as more gooder.

Add option to CLI tool to specify VIIRS APID order. **NOT TESTED

Replaced the merge code with something less efficient but easier to
grok.

New stream impl that does not read the data that will be used to create
and index used in new merge code.

* Ensure streams is a know iterable type (deque)
* Make sure to reuse a stream in the case of a bad packet group. This was
  done with an exception so it was handled the same way as the other 2
  major cases.

bug fix, we were doing streams.remove(stream) inside a "for stream in streams:" which gave incorrect results

Fixes issue #1.