`dump-things` needs multi-record file storage

[mih]

We can easily hit file system limits for some record types (all files relevant to anything).

If the individual records are small, it can make sense to compile multiple records into a single file. In YAML/JSON, the first layer of an object could be a record ID.

A dedicated set of ID functions should support that. Maybe IDs with a colon delimiter. Before the colon is the file name, after it there record ID.

This has trade-offs, of course. More data needs to be read, before the actual record is found. Nothing is perfect.

[mih]

I transferred this issue here from the concepts repo. We di habe sqlite now, which is multirecord storage.

Still, it is worth thinking about other integrations. For the various "seed repos" that hold records to bootstrap sites, I use a json lines file per type, for example.