Fix sandbox cmd timeout by enforcing SDK request timeout (not payload field)

[Copilot]

cmd exposed a timeout parameter in Python SDK, but the underlying /processes/cmd API does not define that field, so timeout was effectively not enforced. This change moves timeout handling to the SDK transport layer to provide actual client-side truncation semantics.

• Problem alignment

  • Removed reliance on unsupported timeout in /processes/cmd request body.
  • Preserved cmd(..., timeout=...) API surface while making it effective.

• Implementation

  • Updated sandbox data API templates:
    • agentrun/sandbox/api/__aio_data_async_template.py
    • agentrun/sandbox/api/__code_interpreter_data_async_template.py
  • Regenerated SDK outputs:
    • agentrun/sandbox/api/aio_data.py
    • agentrun/sandbox/api/code_interpreter_data.py
  • cmd/cmd_async now pass Config(timeout=timeout) into post/post_async (config=), instead of serializing timeout into request body.

• Behavioral change

  • timeout now controls HTTP request timeout at the SDK client layer.
  • timeout=None keeps previous “no explicit cmd timeout override” behavior (no per-call timeout config injected).

# before: timeout sent in body (API does not define it)
api.post("/processes/cmd", data={"command": cmd, "cwd": cwd, "timeout": 30})

# after: timeout enforced by SDK request config
api.post(
    "/processes/cmd",
    data={"command": cmd, "cwd": cwd},
    config=Config(timeout=30),
)