Zero-shot and Few-shot Learning with Instruction-following LLMs for Claim Matching in Automated Fact-checking