Energy-Efficient Deployment of Deep Learning Applications on Cortex-M based Microcontrollers using Deep Compression