Fix #379: Regression with Any and required

[CedricCabessa]

In 0.9.3, the Schemas associated to the Any validators where setup with the
required argument given to Any

https://github.com/alecthomas/voluptuous/blob/0.9.3/voluptuous/validators.py#L218

This allow us to write something like:

Schema(Any({'a': int}, {'b': str}, required=True))

In recent version, the required keyword is ignored.

This patch pass the required argument to the Schemas associated with the
SubValidator.

[coveralls]

Coverage increased (+0.008%) to 95.396% when pulling cf3e2dc on CedricCabessa:fix-379-regression-with-Any-and-required into 7e18e1c on alecthomas:master.

[coveralls]

Coverage increased (+0.02%) to 95.404% when pulling 7b4efb4 on CedricCabessa:fix-379-regression-with-Any-and-required into 7e18e1c on alecthomas:master.

[alecthomas]

Aaah, nice find. Thank you for the PR. Just one minor change and I'm happy to merge.

[CedricCabessa]

I just amend my commit with your remark. Thanks

However, I just found out that some case do not work :-(

like "dict in dict", the empty dict is considered valid but should not

schema = Schema(Any({'a': Any(float, int)},
                         {'b': int},
                         {'c' : {'aa': int}},
                         required=True))

If it's ok, I'll try to come back to you with another commit on this branch to fix that.
Or you can merge this and I'll open another PR, as you prefer

[CedricCabessa]

Here is the 2nd part of the patch to allow "recursive schema"
Let me know what you think :-)

[alecthomas]

Is this still a leftover? Should this be v.required = self.required?

[CedricCabessa]

No, it is correct. Test fail otherwise

Also note that the line must be inside the loop. I'm not totally clear on what happen... but tests are passing ...

[alecthomas]

Thanks!