Add ability to decode across multiple frames

[gdt]

There is at least one device (see #2314) where one can only decode a pair of messages jointly, meaning that decoding each indivdually and having some other process assemble the parts is infeasible. This issue is a feature request to enable decoders to do this joint decoding, necessarily involving storing state from one message for joining with a later message. This issue is not about any other kind of joint processing, and in particular it excludes "turn sensor off if new report not received", and anything that could be done in a second program receiving a json feed.

This splits the joint decode feature from #2635.