[vLLM] vLLM ResponsesAPI & Tool Calling H1 2026 lookahead

### 🚀 The feature, motivation and pitch ### looking back on the past few months vLLM has made a lot of progress on responsesAPI & tool calling in the past few months. To summarize, we've implemented the following: - MCP support for GPTOSS: after GPT-OSS was release in August 2025, we added a series of PRs to support running GPTOSS with MCP: https://github.com/vllm-project/vllm/blob/main/vllm/entrypoints/openai/responses/context.py#L663 - MCP support for all models: https://github.com/vllm-project/vllm/issues/30115. We added the ability to run MCP for _all_ models through a ParsableContext. - Function tool calling: https://github.com/vllm-project/vllm/pull/26874 - We've also added a series of improvements to vLLM responsesAPI, such as fixing tool parsing issues: https://github.com/vllm-project/vllm/pull/30304, supporting partial message generation https://github.com/vllm-project/vllm/pull/32100, adding debugging support with input / output tokens https://github.com/vllm-project/vll