Design for Adding Client State to the Font

As discussed in issue #123 <https://github.com/w3c/IFT/issues/123> we
should consider moving the client state for patch subset to be stored as a
table in the client's font subset. I wrote up a short design doc that
explores what that change would look like in the spec:

Storing Client State in the Font
<https://docs.google.com/document/d/1uws2h-48FRxFQPXN8ZHlmWRl35wRsCREUi_vr7-m9cY/edit?usp=sharing>

In my opinion this looks like a net benefit to the protocol/specification
so I'm in favour of making this change. However, I'd like to hear the
opinions of others before going ahead with the change. So please review
and let me know your thoughts. We could also plan to discuss this at the
next meeting.