Improving the infrastructure around "names of symbols"

[mshinwell]

Please see description in rfcs/symbols.md.

(Rendered)

[gasche]

My impression from the dev-meeting discussion we had was that there was a sort of consensus on leaving the backend out for a first iteration, and concentrating on a symbol abstraction for the middle-end -- if I understand correctly this is the part of this work that was most useful for other to build on, for example #11.