Add support for gpt-4o-mini in ModelType

[realdavidvega]

Background

ModelType didn't have the enum of gpt-4o-mini with the right encoding configuration, leading to falling back to ModelType.TODO, which has CL100K encoding.

This led to errors when using the tokenizer decoder, like in the AI.runWithEnumSingleTokenSerializer, which was used for one token enums.

This PR

Adds ModelType.GPT_4O_MINI with O200K_BASE encoding and 128000 context length.

Example:

@Serializable
enum class Sentiment {
  positive,
  negative,
  neutral
}

suspend fun main() {
  val sentiment = AI<Sentiment>("I love Xef!", config = AIConfig(model = CreateChatCompletionRequestModel.gpt_4o_mini))
  println(sentiment) // positive
}

which returns positive.

[Montagon]

🚀