clarify coordinate space versioning

It seems currently that the coord space versioning is stricter than the template versioning - a change in origin could be a new version of a template but forces a new instance of a coord space. I would argue that the template criteria are the right ones for both cases, and that multiple templates should also be allowed to link to a single coordinate space.

Specifically, I would say that rigid transformations of the space should be allowed as different versions - this would allow the 2020 CCF space to be just another version. A new coordinate space instance would be required when:
- Species or age group changes
- Sample preparation changes significantly (ie with nonlinear effects)
A new template would be required under either of these conditions, and also for new imaging modalities.

If we implement this, the template manifest.json (which seems to exist but isn't documented) should link to the relevant coordinate space. The existing link from the coord space to its "primary" template should stay, but future templates can be associated with an existing space by this approach.