To avoid conflicts with the build-in `-v` `--version` argument, the
short form of `--verbose` is now `-V`.

If we're not going to be printing the formatted data in the log, don't
bother calculating it.  On my system, this speeds up the loading of
polar-graphs-demo.pxp by a factor of 10.

Between binarywave and record.variables, there was a good deal of
parsing that was conditional on previously parsed data.  Rather than
continue writing spaghetti code to handle each specific case, I've
taken a stab at a general framework for updating structures during
parsing (DynamicStructure and DynamicField).  The docstrings should
explain how they work.  The implementation still has a few holes, but
it works on each of the files in my test suite.

This makes more sense than storing them as a tuple.

Also, use wave_record.wave_info['bname'] as the name for storing the
wave in the filesystem view returned by packed.load.

We'll be needing it for packed.load, so it's not binarywave-specific.

Also moved byte_order to the util module and renamed a few functions
internal to the binarywave module so they start with an underscore.

The loadibw -> load (and saveibw -> save) change is because

  from igor.binarywave import load

gives you enough of an idea about what you're importing.  If you want
to keep it explicit in your client module, use

  from igor.binarywave import load as loadibw

which we do in the test.py module.

Samples are from TN003.zip, downloaded from
  ftp://www.wavemetrics.net/IgorPro/Technical_Notes/TN003.zip

I'm confident in the bytes -> strings conversion, but not entirely
sure how the strings get mapped into multidimensional arrays.

Otherwise an earier switch to a non-native byte ordering will confuse
the current load.

It's a utility for the MFP3D driver, not a driver in its own right.