Zero-shot Domain-sensitive Speech Recognition with Prompt-conditioning Fine-tuning