fix: recompute Copilot api_mode after model switch

Recomputes GitHub Copilot api_mode from the selected model in the
shared /model switch path.  Before this change, Copilot could carry a
stale codex_responses mode forward from a GPT-5 selection into a later
Claude model switch, causing unsupported_api_for_model errors.

Cherry-picked from #10533 by @helix4u with:
- Comment specificity (Provider-specific → Copilot api_mode override)
- Fix pre-existing duplicate opencode-go in set literal
- Extract test mock helper to reduce duplication
- Add GPT-5 → GPT-5 regression test (keeps codex_responses)

hermes_cli/model_switch.py

@@ -457,6 +457,7 @@ def switch_model(
         ModelSwitchResult with all information the caller needs.
     """
     from hermes_cli.models import (
+        copilot_model_api_mode,
         detect_provider_for_model,
         validate_requested_model,
         opencode_model_api_mode,
@@ -714,8 +715,12 @@ def switch_model(
     if validation.get("corrected_model"):
         new_model = validation["corrected_model"]
 
+    # --- Copilot api_mode override ---
+    if target_provider in {"copilot", "github-copilot"}:
+        api_mode = copilot_model_api_mode(new_model, api_key=api_key)
+
     # --- OpenCode api_mode override ---
-    if target_provider in {"opencode-zen", "opencode-go", "opencode", "opencode-go"}:
+    if target_provider in {"opencode-zen", "opencode-go", "opencode"}:
         api_mode = opencode_model_api_mode(target_provider, new_model)
 
     # --- Determine api_mode if not already set ---
@@ -1098,5 +1103,3 @@ def list_authenticated_providers(
     results.sort(key=lambda r: (not r["is_current"], -r["total_models"]))
 
     return results
-
-