DocumentHead without NAVM

[cole-miller]

Currently parsing::document always requires enough bytes be presented to determine whether there's a NAVM chunk, and if one is found, the whole chunk must be presented. An important justification for the whole incremental parsing strategy is that you can learn about a document when you only have the first blob of N bytes (delivered over the wire, say), where N is smallish, instead of waiting for all the data to arrive. Always consuming the NAVM chunk increases that N. I don't know whether the increase is important, and I also don't have a detailed scheme for consuming the NAVM chunk "lazily" yet.

[cole-miller]

Idea: introduce NavmP, and then DocumentHead::MultiPage has Either<NavmP, ComponentP> instead of RawNavm, ComponentP. Implement NavmP::feed with the obvious semantics. There is an edge case where we get enough bytes to parse the DIRM stuff but not enough to determine whether there is a NAVM or not; I think in that case we should just demand more bytes (Progress::None)