feat: add streaming support (both steps and tokens) to the Python REST client + pytest

[cpacker]

This PR adds support for streaming to the Python REST client. See the new test code for an example:

def test_streaming_send_message(client: Union[LocalClient, RESTClient], agent: AgentState):
    if isinstance(client, LocalClient):
        pytest.skip("Skipping test_streaming_send_message because LocalClient does not support streaming")
    assert isinstance(client, RESTClient), client

    # First, try streaming just steps
    # TODO

    # Next, try streaming both steps and tokens
    response = client.send_message(
        agent_id=agent.id,
        message="This is a test. Repeat after me: 'banana'",
        role="user",
        stream_steps=True,
        stream_tokens=True,
    )

    # Some manual checks to run
    # 1. Check that there were inner thoughts
    inner_thoughts_exist = False
    # 2. Check that the agent runs `send_message`
    send_message_ran = False
    # 3. Check that we get all the start/stop/end tokens we want
    #    This includes all of the MessageStreamStatus enums
    done_gen = False
    done_step = False
    done = False

    # print(response)
    assert response, "Sending message failed"
    for chunk in response:
        assert isinstance(chunk, MemGPTStreamingResponse)
        if isinstance(chunk, InternalMonologue) and chunk.internal_monologue and chunk.internal_monologue != "":
            inner_thoughts_exist = True
        if isinstance(chunk, FunctionCallMessage) and chunk.function_call and chunk.function_call.name == "send_message":
            send_message_ran = True
        if isinstance(chunk, MessageStreamStatus):
            if chunk == MessageStreamStatus.done:
                assert not done, "Message stream already done"
                done = True
            elif chunk == MessageStreamStatus.done_step:
                assert not done_step, "Message stream already done step"
                done_step = True
            elif chunk == MessageStreamStatus.done_generation:
                assert not done_gen, "Message stream already done generation"
                done_gen = True

• Add support for streaming steps (stream_steps) to Python REST client
• Add support for streaming tokens (stream_tokens) to Python REST client

Note: punted to future PR

• Add support for streaming steps (stream_steps) to Python local client
• Add support for streaming tokens (stream_tokens) to Python local client

Add streaming tests to the client:

pytest tests/test_client.py::test_streaming_send_message

[sarahwooders]

lgtm!