This should help protect against faulty migrations.

Fixes #655

With a private password this time.

This involves a change to the CSV format to allow a type to be specified
in the first column.

Fixes #553.

This was causing tox to choke as there were conflicting requirements
(well, kind of).  Removing it allows tox to run normally but means we
can't run 'upgrade' without conflict.

I've adjusted the makefile to use install for the 'ci' target now
instead of upgrade.

Fixes #600

* Doesn't use progressive plugin on travis (which made the console
  output unreadable).
* Allows arbitrary nose options to be used with ./runtests.py.  Before,
  they might get clobbered.
* Avoid dupliation issue with nose picking up options independentally
  from the testrunner script

Default behaviour is still the same.

The new make docs command also opened the built index.html file.
Unfortunately, there's no cross-platform compatible way to do this apart
from checking for the existence of xdg-open, start, and open... that's a
bit much for a convenience feature.

Installing the requirements.txt after running 'setup.py develop' can
cause problems as it doesn't respect the version ranges from setup.py.
This was picked up by Hudson which was installing Django==1.5 when
upgrading django-nose.

This is addresses by:

* Adding Django to requirements.txt
* Running 'setup.py develop' AFTER installing requirements.txt

Pulling this out of .travis.yaml as it was causing problems with the
django version testing and didn't seem to be the documented way of using
the 'env' setting.

Reworks a commit from #578

Was getting pretty bored of creating one everytime I want to test the
offers dashboard.

Fixture is loaded as part of 'make sandbox|demo' targets.

Adds support for tox to test locally.
Adds support for Travis.

Seems nicer than the generate_css.sh script there before.

It seems to get confused by .pyc files.

The project author sanctions this work-around:
http://stackoverflow.com/questions/2386975/no-source-for-code-message-in-coverage-py

'yolk -U' is very useful...

* Remove accidentally commmitted module code
* Adjust gitignore files
* Add MakeFile target for pulling down puppet modules
* Add settings file for MySQL/Postgres

When DB didn't exist, the rm line would exit 1 causing make to bail out.
Using 'true' solves this although I'm sure there's a better way of doing
this.  It's good enough for now.

This is to ensure that pull requests fail when they don't obey pep8
enough.

Remove old recreatedb script in favour of using the makefile.