server: observe resumable stream cancellation during first result wait (#27481)

[loulanyue]

Summary

Fixes #27481.

For requests carrying X-Conversation-Id (resumable streams), when an explicit DELETE /v1/stream?conv_id=... arrived during prompt evaluation / before the first result, handle_completions_impl polled req.should_stop instead of res->should_stop(). Because the peer socket was still connected, the cancellation was not observed before the first token, keeping the slot/GPU processing.

Changes

• In tools/server/server-context.cpp (handle_completions_impl), poll res->should_stop() when waiting for the first streaming result and verify first_should_stop() in the assertion. This matches the effective_should_stop predicate already used for subsequent chunks in the stream generator.

[ggml-gh-bot]

Hi @loulanyue, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• PR Template not respected: Please respect the template when creating a new pull request. Make sure to fill out all required sections.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.