Add typed createChatCompletion method for structured response handling

[jasonqu]

Add a typed createChatCompletion method for convenience, which could simplify the OpenAI Json Structured Outputs example to code like this:

// if MathReasoning derives zio.json
val response : MathReasoning = syncClient.createChatCompletion[MathReasoning](chatRequestBody) {
  _.fromJson[MathReasoning]
}

The parseFunction is defined as String => Either[String, T] in order to remind developer to treat the error as kind of "checked exception", and parsed error will be rethrow as DeserializationOpenAIException.

Other parseFunction example:

// uPickle
Try(upickle.read[T](rawJson).toEither.left.map(_.getMessage())
// or simply throw the original exception
Right(upickle.read[T](rawJson))

// circe
parse(rawJson).left.map(_.getMessage())

// zio.json
rawJson.fromJson[T]

So this method could be an idiomatic way to create a "typed chat", but might not be a urgent demand.

[adamw]

Thanks!