Using the non-async.io streaming api with add_timestamps can produce error

[robbie-anam]

When invoking the streaming api via:

output_stream = context.send(
                    model_id=self.model_id,
                    transcript=self.read_from_queue_generator(self.current_turn.metrics_context),
                    voice_id=self.voice_id,
                    output_format=self.output_format,
                    add_timestamps=True,
                )

                for chunk in output_stream:

The following error is thrown:

Traceback (most recent call last):
  File "<omitted>/lib/python3.11/site-packages/cartesia/_websocket.py", line 141, in send
    if response["data"]:
       ~~~~~~~~^^^^^^^^
KeyError: 'data'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/<omitted>", in run
    for chunk in output_stream:
  File "<omitted>/lib/python3.11/site-packages/cartesia/_websocket.py", line 169, in send
    raise RuntimeError(f"Failed to generate audio. {e}")

The code does not seem to be handling the case where "data" is not present. The culprit(s) are these lines of code in _websocket:
if response["data"]

You can demonstrate the issue by running a simple example:

response = {}
if response["data"]:
    print("Everything ok")

The "data" field is not meant to be present in word_timestamps responses.

[robbie-anam]

I've opened a PR with a fix for this here: #19