fix(chat): prevent model changes from aborting active turns

[heath-s]

Summary

• keep the current provider turn streaming when model, effort, or provider selection changes
• apply the new selection to subsequent turns without task-wide cleanup during an active turn
• preserve existing runtime cleanup for idle provider switches
• add regression coverage for active and idle selection changes

Root cause

The model/effort selector re-selected the task provider on every change. setTaskProvider unconditionally called task-wide provider cleanup, which aborted any active turn even when the provider had not changed.

Validation

• bun test tests/task-provider-selection.test.ts tests/task-stream-switch-regression.test.ts tests/prompt-draft-runtime.test.ts tests/model-effort-selector-utils.test.ts tests/model-runtime-preferences.test.ts (25 passed)
• bun run typecheck
• Prettier check for changed files

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b57192156b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".