feat(example): Updated server example (batch processing, /v1/responses api, response parsing)

[abetlen]

This PR adds an updated OpenAI compatible web server that depends only on the low-level C bindings.

Some of the new features this server supports:

• Batched inference scheduling (multiple streaming concurrent requests as well as requests with multiple completions)
  • Add batched inference #771
• Automatic model-type aware prompt caching for attention, recurrent, and hybrid models
• Response parser based on HF response_schema for automatic tool calling with streaming support depending on the schema
  • function_call and tool_choice didn't have "auto" feature implemented #1338
• /v1/responses support (stateless)

Running the server

uv run --script server.py -C config.json

example `config.json`

{
  "server": {
    "host": "0.0.0.0",
    "port": 8000
  },
  "model": {
    "from_pretrained": {
      "repo_id": "lmstudio-community/Qwen3.5-0.8B-GGUF",
      "filename": "Qwen3.5-0.8B-Q8_0.gguf"
    },
    "n_ctx": 512,
    "n_batch": 64,
    "n_ubatch": 64,
    "n_seq_max": 8,
    "threads": 8,
    "threads_batch": 8,
    "prompt_chunk_size": 32,
    "kv_unified": true,
    "response_schema": {
      "type": "object",
      "properties": {
        "role": {
          "const": "assistant"
        },
        "reasoning_content": {
          "type": "string",
          "x-regex": "^(?:<\\|im_start\\|>assistant\\n)?(?:<think>\\n)?(.*?)(?=</think>)"
        },
        "content": {
          "type": "string",
          "x-regex": "^(?:<\\|im_start\\|>assistant\\n)?(?:(?:<think>\\n)?.*?</think>\\s*)?(.*?)(?=\\s*<tool_call>\\n|<\\|im_end\\|>$|$)"
        },
        "tool_calls": {
          "type": "array",
          "x-regex-iterator": "<tool_call>\\n(.*?)\\n</tool_call>",
          "items": {
            "type": "object",
            "properties": {
              "type": {
                "const": "function"
              },
              "function": {
                "type": "object",
                "properties": {
                  "name": {
                    "type": "string",
                    "x-regex": "^<function=([^>\\n]+)>\\n"
                  },
                  "arguments": {
                    "type": "object",
                    "x-regex": "^<function=[^>\\n]+>\\n(.*?)\\n</function>$",
                    "x-regex-key-value": "<parameter=(?P<key>[^>\\n]+)>\\n(?P<value>.*?)\\n</parameter>",
                    "additionalProperties": true
                  }
                },
                "required": [
                  "name",
                  "arguments"
                ]
              }
            },
            "required": [
              "type",
              "function"
            ]
          }
        }
      },
      "required": [
        "role"
      ]
    }
  }
}

---

The purpose of this example is to solve a bunch of modernisation projects we need for this library which will be applied (slowly) and in a way that doesn't break backwards compatibility.