Efficient transformer adaptation for analog in-memory computing via low-rank adapters (Full Report)