[ChatError] Fixed implementation

[pvieito]

This fixes the implementation of #43:

• The param and code error fields are optional.
• We have to check if there is a ChatError before we try to decode the OpenAI response, as the later always succeds due to this change.

[bogdanf]

Very good catch. Basically, the whole error handling is useless without it.

[adamrushy]

Great catch 👏🏼

[azamsharp]

I updated the packages and the decoding error is now gone but when I make a request I get the following:

OpenAI<TextResult>(object: nil, model: nil, choices: nil, usage: nil, data: nil)

openAI.sendCompletion(with: chatText, maxTokens: 500) { result in
            switch result {
                case .success(let success):
                  // OpenAI<TextResult>(object: nil, model: nil, choices: nil, usage: nil, data: nil)
                    print(success) 

The success contains all nil values.