Expose token usage from openAIResponsesAdapter

Is your feature request related to a problem? Please describe. openAIResponsesAdapter() currently ignores response.completed, including response.usage.

Describe the solution you'd like Could the adapter expose an optional callback?

openAIResponsesAdapter({ onUsage: (usage) => { // input_tokens, output_tokens, total_tokens }, }); This would let applications display token usage or estimated remaining context without implementing a custom stream adapter.

Describe alternatives you've considered Local adapter that wraps/forks openAIResponseAdapter

Additional context This would help implement a context available/used counter in the composer