Text this: Balancing Privacy and Utility in Split Learning: An Adversarial Channel Pruning-Based Approach