Testing part marking: manage the link to the runtime with a display object

It's just occurred to me that the part editor does a lot of fiddling about with extracting stuff from the runtime, when it could just implement a PartDisplay object, where the API is fairly well-defined.

I think this would help to separate the runtime interface bits from the rest of the editor. At the moment it's a bit of a mess.

Start by locating the part editor and tracing where it extracts information from the runtime. Review the proposed PartDisplay boundary and the existing runtime-facing interactions before deciding the API. Done means the editor's runtime interface is separated behind a clearly defined display object, with the relevant behavior still working.

Written by the indexing model from the issue text.