Fix missing token usage from some external model endpoints

[jakesteelman-insurica]

This PR will add an extra condition to check for chunks that have token usage information but do not have choices or delta properties. I opened #198, then was able to trace the cause down to GPT-5 sending a chunk at the end that was not handled by any branch of the if/elif statements since it did not have chunk.choices or chunk.delta.

Here's an example of the last chunk from calling llm.stream("..."), if you add a print('Chunk Received:', chunk) statement on line 635 of chat_models.py

from databricks-langchain import ChatDatabricks

llm = ChatDatabricks(endpoint="gpt-5")
llm.stream("hello")

Chunk Received: ChatCompletionChunk(id='chatcmpl-CO5Nwd3B1mZzhy32nob4TgRGSOYnY', choices=[Choice(delta=ChoiceDelta(content='Hello', function_call=None, refusal=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1759856608, model='gpt-5-2025-08-07', object='chat.completion.chunk', service_tier='default', system_fingerprint=None, usage=None, obfuscation='beO2Lti')
Chunk Received: ChatCompletionChunk(id='chatcmpl-CO5Nwd3B1mZzhy32nob4TgRGSOYnY', choices=[Choice(delta=ChoiceDelta(content=' ,', function_call=None, refusal=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1759856608, model='gpt-5-2025-08-07', object='chat.completion.chunk', service_tier='default', system_fingerprint=None, usage=None, obfuscation='c5xtxoC')
...
Chunk Received: ChatCompletionChunk(id='chatcmpl-CO5Nwd3B1mZzhy32nob4TgRGSOYnY', choices=[], created=1759856608, model='gpt-5-2025-08-07', object='chat.completion.chunk', service_tier='default', system_fingerprint=None, usage=CompletionUsage(completion_tokens=267, prompt_tokens=4861, total_tokens=5128, completion_tokens_details=CompletionTokensDetails(accepted_prediction_tokens=0, audio_tokens=0, reasoning_tokens=192, rejected_prediction_tokens=0), prompt_tokens_details=PromptTokensDetails(audio_tokens=0, cached_tokens=0)), obfuscation='wts2')

[WIll-Xu35]

Facing the same issue here

[sunishsheth2009]

Thank you for your contribution here and helping to fix this issue

[sunishsheth2009]

Can we add some test here: integrations/langchain/tests/unit_tests/test_chat_models.py for this where the last chunk is usage data and those are added to the response correctly?