Text this: Hybrid LSTM–Attention and CNN Model for Enhanced Speech Emotion Recognition