### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

When using the OpenAI Responses API (`/v1/responses`) through LiteLLM, Langfuse traces show `input: 0`, `output: 0`, `total: 0` token usage despite the underlying model returning correct token counts.

This is because the Langfuse callback reads `prompt_tokens`/`completion_tokens` from the usage object, but Responses API responses use `input_tokens`/`output_tokens` instead. The callback never falls back to the Responses API field names, so it logs zeros.

Additionally, the response output content for Responses API calls was not being logged — the callback only extracted content from Chat Completion-style response objects, not from `ResponseCompletedEvent.response.output`.

### Relevant log output

```shell # Trace in Langfuse shows: input: 0, output: 0, total: 0

# But LiteLLM spend logs show correct counts: input_tokens: 16, output_tokens: 21 ```

### Are you a ML Ops Team?

No, individual self-hosted user.

### What LiteLLM version are you on?

v1.74 (latest at time of reporting)

### Relevant stack trace

```python # In langfuse.py _log_lang…