fix(station): base AllowListed on independently listed addresses

[MRmarioruci]

Change

RequestPolicyRule::AllowListed treats a destination's presence in the address book as approval for a transfer to it. Address book writes and an account's transfer policy are configured independently and commonly sit at different approval tiers, so presence on its own is not necessarily an independent signal about the destination.

• AddressBookEntry gains last_modified_by: Option<UUID>, set on create and on edit. Edits are covered as well as creation, since changing an existing entry reaches the same state.
• AllowListed looks the entry up rather than only testing existence, and does not treat it as independent when last_modified_by matches the transfer's requested_by. Listing an address and spending to it therefore involve two different people.
• The requester flows in from request.requested_by in the add/edit request executors.

Backwards compatibility

Entries stored before the field existed decode with last_modified_by: None and are still honoured, so existing allow-lists keep working across the upgrade.

This is asserted rather than assumed: decodes_entries_stored_before_last_modified_by_existed encodes a struct with the pre-change shape using the same CBOR serializer #[storable] uses, decodes it as the current AddressBookEntry, and checks the field defaults to None with every other field intact.

Tests

• rejects_an_address_the_requester_listed_themselves
• approves_an_address_listed_by_someone_else — the feature still works
• approves_entries_that_predate_authorship_tracking — no regression for existing allow-lists
• decodes_entries_stored_before_last_modified_by_existed — the compatibility proof above

cargo test -p station --lib passes (398). Clippy warning count unchanged from baseline; cargo fmt clean.

Unrelated observation

migration_tests::make_repository_snapshots only writes the .bin snapshots; nothing reads them back or asserts against them. They are regenerated artifacts rather than a compatibility check, which is why they show dirty after any test run, and why the compatibility test above stands alone. The address_book_repository_v3.bin change here is just the regenerated artifact reflecting the new field.