Add Tokenizer functionality to API

Having access to the models tokenizer is extremely useful for counting tokens, and managing the context window. In a lot of cases its essential to get an LLM implementation to work properly. The model already has the tokenizer loaded, and ollama's backend, llama.cpp, already has an interface for the tokenizer, so it shouldn't be that difficult to implement into the API. Unless this is already a functionality on the API, in which case I'm sorry, but I just didn't see it in the documentation.