Leverage copy-on-write trees for data recovery

[Nuhvi]

Since we will be using Merkle Treap for efficient verification and sync, we might as well use their copy-on-write nature, for versioning and data recovery.

There are two main scenarios where a user needs to recover data:

1. Error
2. Malicious homeserver

To recover from mistakes, homeservers can offer to keep up to n snapshots of each repo/drive as long as it fits within the user storage quota.

To recover from Malicious homeservers deleting all users data, there must be some mirror or archive pulling snapshots from the homeserver, however, it can't keep infinite amount of history, so the homeserver can create too many malicious versions of the data, that the mirrors drop all the older valid versions.

To avoid that, mirrors should sync a maximum number of snapshots from a homeserver, for example by only syncing once an hour and only getting the most recent version.

This way, mirrors offer users m number of days of history to recover within. Hopefully a user can notice their homeserver malice, and react quickly to migrate to another homeserver, or download their entire data if it is small enough to fit in their personal computer.

[Nuhvi]

Another way to leverage snapshots is; while migrating to another homeserver, a user can specify a specific snapshot from the old homeserver (usually the current one) for the new homeserver to keep no matter what the old homeserver does during migration, so once that snapshot is downloaded, the user is satisfied that this is enough to switch homeservers.