Ensure Maybe propagates error information

[monopolis]

Due to Any evaluating all options, and raising the exception
from the first encountered error, Maybe would discard the
expected error raised by the provided validator.

This commit changes the order of the validators in Maybe so
that the first evaluated error (and thus returned) is that of
the provided validator.

[coveralls]

Coverage increased (+0.8%) to 95.617% when pulling b4e3e8d on monopolis:issue-369 into 7151438 on alecthomas:master.

[svisser]

Hmm, it's interesting because the test case now fails in Python 3.x due to:

  File "/home/travis/build/alecthomas/voluptuous/voluptuous/validators.py", line 589, in __call__
    if self.min is not None and not v >= self.min:
TypeError: '>=' not supported between instances of 'NoneType' and 'int'

This previously didn't happen as it checked for None first.

[monopolis]

Please see pull request #414 for a potential fix for the incomparable values. It should solve this issue at the very least.

My guess is that there probably are other validators that should check for TypeError's as well - e.g. Length comes to mind.

[spacegaier]

Please see pull request #414 for a potential fix for the incomparable values. It should solve this issue at the very least.

My guess is that there probably are other validators that should check for TypeError's as well - e.g. Length comes to mind.

@svisser The mentioned PR #414 has by now been merged and indeed fixes this issue for me.
@monopolis I added additional TypeErrors to Clamp and Length, so that is covered now.