add documentation for rich inputs

[exarkun]

The "rich input" and "input context" ideas in machinist aren't implementations of a very common or widely used concept in the "finite state machine" sphere (though they are certainly inspired by ideas I saw in other libraries). The current documentation doesn't really explain them well enough for anyone to know how they work or why they're useful.

Add some docs explaining them. Probably with an example or two.

[exarkun]

There is some overlap here with #3, likely, which also discusses an example for rich inputs. There's probably room for more than one - but examples aren't the only kind of documentation machinist should have. A little prose that explains why the heck there is even such a thing as "rich inputs" would be pretty handy.