`write_dm_with_retry` and `write_room_message_with_retry` are identical retry loops

These two functions in `persistance_actor/handlers.rs` share the exact same structure: clone client, loop with attempts, sleep with linear backoff, return last error. The only difference is which gRPC method they call.

hint: generics, Both callers become one-liners.

Start in persistance_actor/handlers.rs by reading write_dm_with_retry and write_room_message_with_retry, then inspect their callers. Compare the retry attempts, linear backoff, client cloning, and error handling; done means the shared generic retry behavior is preserved while both callers become one-liners.

Written by the indexing model from the issue text.