Partially fix UTF-8 and PSON encoding tests for Ruby 1.9

With the introduction of real encoding support in Ruby 1.9, various problems
including incorrectly encoded UTF-8 strings crop up.

By flagging the encoding of the spec file to Ruby, and forcing binary encoding
on strings we compare that contain invalid UTF-8, we can bypass some of those
headaches.

Since PSON is, by design, not compatible with JSON and can do useful things
with arbitrary binary data, this is the most reasonably current choice for how
to handle this troublesome set of data.

Signed-off-by: Daniel Pittman <daniel@puppetlabs.com>