Support origin pagination without origin ids

in origin_get_range and origin_search.

URL hashes would be the obvious solution as pagination index, but that wouldn't work efficiently with Cassandra (although it's possible).

I propose we move to opaque tokens.

(This could also be extended to all other endpoints in the future, as each currently has its own scheme and argument names for pagination.)

We also need to find a name for this token, as the same mechanism will be used in multiple components and endpoints, eg:

• cursor (name conflict with SQL cursors)
• continuation token
• scroll token
• ?

---

Migrated from T1912 (view on Phabricator)