fix: add proper timeout error; improve backoff logic on replication recovery

[filipecabaco]

What kind of change does this PR introduce?

• Properly log out the timeout error on replication connection
• Improve backoff policy to include exponential backoff to provide more time to db instance to recover

[edgurgel]

Would you mind a comment on what's the expected growth on the backoff here?

[blacksmith-sh]

Found 2 test failures on Blacksmith runners:

Failures

Test	View Logs
test WAL bloat handling track PID changes during WAL bloat creation (Realtime.Integrati on.RtChannelTest) › Parameters: %{serializer: Phoenix.Socket.V1.JSONSerializer} › test/ integration/ rt_channel_test.exs:2301 › ** (MatchError) no match of right hand side value: %Postgrex .Result{command: :select, columns: ["active_pid"], rows: [], num_rows: 0, connection_id : 1151, messages: []} › code: %{rows: [[new_db_pid]]} = Postgrex.query!(db_conn, active _slot_query, [])/stacktrace	View Logs
test WAL bloat handling track PID changes during WAL bloat creation (Realtime.Integrati on.RtChannelTest) › Parameters: %{serializer: RealtimeWeb.Socket.V2Serializer} › test/ integration/ rt_channel_test.exs:2301 › ** (MatchError) no match of right hand side value: %Postgrex .Result{command: :select, columns: ["active_pid"], rows: [], num_rows: 0, connection_id : 1533, messages: []} › code: %{rows: [[new_db_pid]]} = Postgrex.query!(db_conn, active _slot_query, [])/stacktrace	View Logs

[edgurgel]

So if we got num_rows different than 0 but this Connect proces does not own the ReplicationConnection process what does this mean in the end?