fix(core): default provider header and chunk timeouts to five minutes

[rekram1-node]

Sets default timeouts of five minutes (300,000 ms) for provider requests: one for waiting on response headers, and one for the gap between streamed response chunks. Requests that exceed either limit fail with a transport error and follow the existing retry policy. The chunk timer resets whenever data arrives, so it limits inactivity, not total response duration.

Both timeouts apply to native @opencode-ai/ai routes and to the AI SDK path. They are carried on HttpOptions so route, model, and request values merge with the usual precedence.

To change or disable them for a provider, set headerTimeout or chunkTimeout under that provider's settings in your project's opencode.jsonc or global ~/.config/opencode/opencode.jsonc:

{
  "providers": {
    "anthropic": {
      "settings": {
        "headerTimeout": 600000,
        "chunkTimeout": false
      }
    }
  }
}

Replace anthropic with your provider. Use a positive number of milliseconds to change a limit, or false to disable it. Omit a setting to use the five-minute default.