Adds support for --test to cli utils

Adds support for --test to cli utils

This branch bundles two related changes, support for --test on relation-get, unit-get and config-get.
This should be similar to the additions made in the golang version. 

Also this changes to always use a formatter for output and the smart output formatter was changed to
YAML dump collection types, normal string values are still printed w/o serialization.

https://code.launchpad.net/~bcsaller/juju/sane_output_test_option/+merge/106067

(do not edit description out of merge proposal)

[bcsaller, 2012-05-16 22:40:34]

Please take a look.

[niemeyer, 2012-05-17 02:42:44]

Some details here are not clear to me. I'd appreciate having a chat about this
feature with you before this goes in.

https://codereview.appspot.com/6216053/diff/1/juju/hooks/cli.py
File juju/hooks/cli.py (right):

https://codereview.appspot.com/6216053/diff/1/juju/hooks/cli.py#newcode228
juju/hooks/cli.py:228: # yaml dump container types
What was your thinking here?  I believe it should dump yaml in all cases. i.e.
"true", not "True".

https://codereview.appspot.com/6216053/diff/1/juju/hooks/commands.py
File juju/hooks/commands.py (right):

https://codereview.appspot.com/6216053/diff/1/juju/hooks/commands.py#newcode80
juju/hooks/commands.py:80: v = json.dumps(v)
json?

[hazmat, 2012-05-18 17:25:47]

bug in base formatter.

https://codereview.appspot.com/6216053/diff/1/juju/hooks/cli.py
File juju/hooks/cli.py (right):

https://codereview.appspot.com/6216053/diff/1/juju/hooks/cli.py#newcode220
juju/hooks/cli.py:220: formatter(result, stream)
None is not a callable type, this tries a double output if formatter is None.
please default formatter to the correct default output.

[bcsaller, 2012-07-06 20:36:07]

Please take a look.

https://codereview.appspot.com/6216053/diff/1/juju/hooks/cli.py
File juju/hooks/cli.py (right):

https://codereview.appspot.com/6216053/diff/1/juju/hooks/cli.py#newcode228
juju/hooks/cli.py:228: # yaml dump container types
On 2012/05/17 02:42:44, niemeyer wrote:
> What was your thinking here?  I believe it should dump yaml in all cases. i.e.
> "true", not "True".

You're correct that non-string types should appear here.

[hazmat, 2012-07-10 22:17:19]

I'm still a little unclear as to the desired behavior outcome and its testing
with this branch, as it currently tested it doesn't seem particularly useful.

https://codereview.appspot.com/6216053/diff/6001/juju/hooks/commands.py
File juju/hooks/commands.py (right):

https://codereview.appspot.com/6216053/diff/6001/juju/hooks/commands.py#newco...
juju/hooks/commands.py:18: class TestOptionMixin(object):
missing doc string describing purpose/usage.

https://codereview.appspot.com/6216053/diff/6001/juju/hooks/commands.py#newco...
juju/hooks/commands.py:34: self.handle_test(result)
there's a lot of unesc. method delegation here. just inline the
test_true/handle_test behavior.

The use of the mixin, also feels like its adding complexity here, around
subclass ordering and proper lineage to customize_parser vs. render (which have
inverse call lineages). Incorporating into the base class might be cleaner via
class flag switch.

https://codereview.appspot.com/6216053/diff/6001/juju/hooks/commands.py#newco...
juju/hooks/commands.py:114: def render(self, result):
Why is this returning None, why is it implemented at all? ditto for the two
other implementations of the same here.

https://codereview.appspot.com/6216053/diff/6001/juju/hooks/tests/test_invoke...
File juju/hooks/tests/test_invoker.py (right):

https://codereview.appspot.com/6216053/diff/6001/juju/hooks/tests/test_invoke...
juju/hooks/tests/test_invoker.py:356: 
this is testing something very different then what the cli usage would commonly
be, namely the cli usage would be testing for values unset.

For values set, a null value seems valid, has the go team documented the
behavior of this cli param?