Consider storing more data in cursor

[Nutomic]

At the moment our db cursors only store the prefix (type), item id and whether it is for previous or next page. With the item id we read the actual item (eg Post) from the database. Then we use fields like post.id, post.published_at etc from the loaded item to sort rows for pagination.

Instead it would be possible to get rid of this extra db read and store the relevant data (id, published_at, etc) directly inside the cursor. The easiest way would be to store the entire Post inside the cursor but this would make cursors too large. Instead the cursor should only contain which are used in .then_order_by() calls. The question is how to keep these two in sync when changes are made over time.

There is no breaking change required as the format of pagination cursors can change across minor versions. So this can be implemented anytime after 1.0

[dessalines]

I spose we could base64 encode the entire post, but that'd be a long cursor.

[dullbananas]

Previously discussed in #4982