Committing the work I've been doing over the past little while on a
rewrite for python-fitparse.

Major Changes
-------------

* New, hopefully cleaner public API with a clear division between
  accessible and internal parts. (Still unstable and partially
  complete.)

* Proper documentation! Going to use Sphinx.

* Unit tests and example programs.

* (WIP) Command line tools (eg a `.FIT` to `.CSV` converter).

* Component fields and compressed timestamp headers now supported
  and not just an afterthought. Closes issues #6 and #7.

* FIT file parsing is generic enough to support all types. Going to
  have specific `FitFile` subclasses for more popular file types
  like activities.

* (WIP) Converting field types to normalized values (for example,
  `bool`, `date_time`, etc) done in a consistent way, that's easy to
  customize by subclassing the converter class. I'm going to use
  something like the Django form-style `convert_<field name>` idiom
  on this class.

* The FIT profile is its own complete python module, rather than
  using `profile.def`.

* Bonus! The profile generation script is less ugly (but still an
  atrocity) and supports every ANT FIT SDK from version 1.00 up to
  5.10.

* A working `setup.py` module. Closes issue #2, finally! I'll
  upload the package to PyPI when it's done.

* Support for parsing one record at a time. This can be done using
  `<FitFile>.parse_one()` for now, but I'm not sure of the exact
  implementation yet.