Bookmark FFI for iOS and Kotlin

[thomcc]

This differs from the doc a bit (The operations in the doc are fairly low level, and also are not how our FFI works, so I took the doc as a "these are the things we need to support" list), but supports all the same operations and is a bit more general (e.g. move is part of update, getting the list of child GUIDs is part of the API to get a bookmark entry in general), and a lot higher level.

I'm completely willing to add the operations from the doc explicitly (implemented on top of what I have) as a further feature. This way was clearer to implement given what we had already, and mor.

The documentation in the Kotlin and Swift files is fairly thorough.

This PR is mostly complete, but still needs:

• Integration with iOS megazord
• Finish documenting how to integrate protobufs on iOS so the next person who has to do it doesn't have to figure out so much from scratch again like I did for this PR.
• Document which methods can throw which errors.
• More test coverage.
• Changelog entry.

I should probably clean up the commit history a bit too, so that we don't have to squash-merge it. Nah lets just squash merge this rather than waste time cleaning up a history that nobody really cares about.

Fixes #727

r? @justindarc for the swift
r? @grigoryk for the kotlin
r? @linacambridge / @mhammond for the rust

Followups:

• BookmarkTreeNode's GUID should probably not be optional.
• Consider implementing fetch_bookmark using fetch_tree when get_direct_children is true
• fetch_tree should return position / parent GUID info so we don't have to fix it up in fetch_public_tree (which requires a query).
• run_maintenance should fix up bookmark tree corruption.
• Change update_bookmark_from_message to pass a raw bookmark into update_bookmark_in_tx (which should be changed / another function added to accept an already-queried raw bookmark)
• Too many bookmark node types!

[thomcc]

r?@fzzzy since he's been reading a bunch about the FFI and seeing how it works in practice might be helpful.

[garvankeeley]

Not a reviewer, but lgtm! and dropped in some drive-by food-for-thought comments.

[justindarc]

LGTM! Some nits about naming, etc.

[linabutler]

This is excellent, @thomcc, thanks! I had mostly nits, and a suggestion for avoiding multiple reads to fetch direct children, but we can tackle those in follow-ups if you'd like. I wouldn't mind reading through this again before it lands, but I think it's in great shape. 🚀

[grigoryk]

I've only looked at the Kotlin APIs, from the point of view of a consuming application.

[grigoryk]

nit: (unsure yet how this might fit w/ the rest of the api) You could enforce the "can't be null for non-root nodes" limitation at the interface level.

[thomcc]

I'd rather keep things simple than optimize for enforcing as much as possible, but if you have a concrete idea then LMK and I'll think about it.

[mhammond]

This looks great Thom and obviously took a heap of work - well done.

Mainly nits, and I'm not looking forward to resolving the conflicts with the bookmark-merge branch, but +1 from me!

[linabutler]

Let's 🚢 this! We can tackle anything that's left in follow-ups. Love the thorough docs, too!