tests: Miscellaneous improvements.

- Add "check" to a lot of commands, to ensure that they succeed.

- Dump flows to files instead of to stdout, to make them easier to
  compare between different runs.

- Compare all the packets that should be sent for a given run, rather
  than just comparing them for a single interface.  This gives a better
  overview on failure (did everything fail? or just one packet for a
  given interface?)

- Better comments and log messages.

Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Numan Siddique <numans@ovn.org>