OpenAI: Count tokens without sending a request to API

There is a Python tool called `tiktoken`, which is an official tokenizer that counts the number of tokens in messages to be sent. It allows you to identify token overflows without sending a request.

If the same official library exists but in Kotlin, we can utilize it; it will speed up the test generation workflow, as there will be no need for HTTP request calls if we already know that the request will fail.

**See:** https://platform.openai.com/docs/guides/text-generation/managing-tokens