Use model catalog collaboration mode messages (#36351)

## What changed

- Add optional `default` and `plan` collaboration-mode messages to model catalog metadata.
- Prefer the active mode's catalog message, including an explicit empty value, while falling back to the existing developer instructions when the catalog variant is absent.
- Track the model in collaboration-mode world state so mode and model changes append the appropriate instructions, and clear prior instructions when no current message exists.
- Preserve catalog collaboration messages when instruction overrides remove other model instruction fields.

## Testing

- Cover catalog deserialization, variant selection, fallback and empty-message behavior, legacy snapshot migration, mode changes, and model changes.

GitOrigin-RevId: d4c6c0028423bc50b0e8175e55586f3e7bc1c3a8

codex-rs/protocol/src/openai_models.rs

@@ -507,6 +507,7 @@ pub struct ModelMessages {
     pub instructions_template: Option<String>,
     pub instructions_variables: Option<ModelInstructionsVariables>,
     pub approvals: Option<ApprovalMessages>,
+    pub collaboration_modes: Option<CollaborationModeMessages>,
     pub auto_review: Option<AutoReviewMessages>,
     pub permissions: Option<PermissionMessages>,
     #[serde(default, skip_serializing_if = "Option::is_none")]
@@ -531,6 +532,12 @@ pub struct ApprovalMessages {
     pub unless_trusted: Option<String>,
 }
 
+#[derive(Debug, Serialize, Deserialize, Clone, PartialEq, Eq, TS, JsonSchema)]
+pub struct CollaborationModeMessages {
+    pub default: Option<String>,
+    pub plan: Option<String>,
+}
+
 #[derive(Debug, Serialize, Deserialize, Clone, PartialEq, Eq, TS, JsonSchema)]
 pub struct AutoReviewMessages {
     pub policy: Option<String>,
@@ -763,13 +770,23 @@ mod tests {
     }
 
     #[test]
-    fn model_messages_deserialize_without_approvals() {
+    fn model_messages_deserialize_without_optional_sections() {
         let messages: ModelMessages =
             from_str(r#"{"instructions_template":null,"instructions_variables":null}"#)
                 .expect("model messages should deserialize");
 
-        assert_eq!(messages.approvals, None);
-        assert_eq!(messages.permissions, None);
+        assert_eq!(
+            messages,
+            ModelMessages {
+                instructions_template: None,
+                instructions_variables: None,
+                approvals: None,
+                collaboration_modes: None,
+                auto_review: None,
+                permissions: None,
+                token_budget: None,
+            }
+        );
     }
 
     #[test]
@@ -860,6 +877,36 @@ mod tests {
         );
     }
 
+    #[test]
+    fn collaboration_mode_messages_preserve_missing_and_empty_values() {
+        let messages: ModelMessages = from_str(
+            r#"{
+                "instructions_template": null,
+                "instructions_variables": null,
+                "collaboration_modes": {
+                    "default": ""
+                }
+            }"#,
+        )
+        .expect("collaboration mode messages should deserialize");
+
+        assert_eq!(
+            messages,
+            ModelMessages {
+                instructions_template: None,
+                instructions_variables: None,
+                approvals: None,
+                collaboration_modes: Some(CollaborationModeMessages {
+                    default: Some(String::new()),
+                    plan: None,
+                }),
+                auto_review: None,
+                permissions: None,
+                token_budget: None,
+            }
+        );
+    }
+
     #[test]
     fn reasoning_effort_accepts_known_and_custom_values() {
         let custom = ReasoningEffort::Custom("future".to_string());
@@ -932,6 +979,7 @@ mod tests {
             instructions_template: Some("Hello {{ personality }}".to_string()),
             instructions_variables: Some(personality_variables()),
             approvals: None,
+            collaboration_modes: None,
             auto_review: None,
             permissions: None,
             token_budget: None,
@@ -952,6 +1000,7 @@ mod tests {
                 personality_pragmatic: None,
             }),
             approvals: None,
+            collaboration_modes: None,
             auto_review: None,
             permissions: None,
             token_budget: None,
@@ -981,6 +1030,7 @@ mod tests {
                 personality_pragmatic: None,
             }),
             approvals: None,
+            collaboration_modes: None,
             auto_review: None,
             permissions: None,
             token_budget: None,
@@ -1013,6 +1063,7 @@ mod tests {
                 personality_pragmatic: None,
             }),
             approvals: None,
+            collaboration_modes: None,
             auto_review: None,
             permissions: None,
             token_budget: None,