How Different Tokenization Algorithms Impact LLMs and Transformer Models for Binary Code Analysis