This commit is in spirit of the previous two and removes unnecessary
duplication of subjects for twts since they can easily be reconstructed
from the parsed elements.

Again, as in the previous commit, there's no need to duplicate the
author of a twt for each and every twt in the C bridge. The Python
binding can take care of that easily itself.

It's sufficient to let the Python wrapper reconstruct the lists from the
parsed elements. No need to pass all these data twice over the C bridge.

I reckon it makes a bit more sense to have the length first and then the
pointer to the very first element of the array than the other way
around.