Handle microseconds in Time SchemaType

[blqn]

Actually, it is impossible to manage microseconds using Time SchemaType:

>>> from colander import Time
>>> from colander import SchemaNode
>>> from datetime import time
>>> node = SchemaNode(Time())
>>> node.deserialize('10:42:42.424242')
...
colander.Invalid: {'': 'Invalid time'}
>>> t = time(10,42,42,424242)
>>> node.serialize(t)
'10:42:42'

This PR handle full ISO format HH:MM[:SS[.mmmmmm]] for deserialization and HH:MM:SS[.mmmmmm] for serialization through time.isoformat.

[digitalresistor]

Instead of trying to guess based upon length, why not just attempts to go through each one in order? Start with the ms one, followed by the one with seconds, followed by the one with just hours:minutes?

Much like the original solution did, but loop over a list of formats.

[blqn]

Because I thought nested exceptions were not very readable and filled the stack trace for not much. While the proposed solution is equivalent and more readable, in my opinion. But maybe I'm missing something and not understanding your proposed loop over formats using exceptions.

[mmerickel]

I would also prefer a solution that looped over the formats and tried them. Finally if none of them worked then raise an exception. You could pass in the final error object as the err key. I don't think any nested exceptions would occur if this were a loop.

I'm concerned about testing the length because I don't actually know if the %H etc always maps to two characters - it defers to libc and possibly any locale settings which may cause unexpected regressions. It's safer to just test the formats and use the one that works since the formats are unambiguous.

[mmerickel]

Loop over the formats instead of relying on length.

[mmerickel]

Thanks @blqn !