Add a vocabulary_size argument to UnicodeCharacterTokenizer

[mattdangerw]

We should add a vocabulary_size argument to the UnicodeCharacterTokenizer layer that clamps the output code point values to be in the range [0, vocabulary_size).

We should also add a vocabulary_size() method as with other tokenizers that returns this value if it has been set, and return None otherwise.

Potential docstring:

        vocabulary_size: Set the vocabulary `vocabulary_size`,
            by clamping all codepoints to the range [0, vocabulary_size).
            Effectively this will make the `vocabulary_size - 1` id the
            the OOV token.

[aflah02]

Hey @mattdangerw
If no one else is interested I can take this up!

[mattdangerw]

Go for it! Thanks!