Created a Reader class to encapsulate reading files with the encoding
and decode_errors attributes.

The spec tests are now excluded from nosetests, by default.  Added
instructions to the README on how to include them.

This brings the spec tests into the nosetest/unittest run.

Currently, `nosetests --with-doctest` (which includes mustache spec
tests) is at the following:

----------------------------------------------------------------------
Ran 234 tests in 0.407s

FAILED (errors=3, failures=27)

All errors/failures are with spec tests.

Adds spec tests.

Double quotes are now escaped, by default (when markupsafe is not available).

This commit removes markupsafe details from the RenderEngine class.

Merry Christmas! :)

Added a unit test and doctest.

We added a case to check that callable return values of __getitem__
are not called.  There is already a test that callable attributes
are called: test_object__attribute_is_callable().

Added a test case for (non-list) "non-false value".

The Renderer class's default loader now inherits the decode_errors
attribute for the instance.

The Renderer class now accepts a loader implementing get()
(e.g. a dictionary or Loader instance) rather than a load_template
function.