Cap realtime mirrored user turns (#17685)

Cap mirrored user text sent to realtime with the existing 300-token turn
budget while preserving the full model turn.

Adds integration coverage for capped realtime mirror payloads.

---------

Co-authored-by: Codex <noreply@openai.com>

codex-rs/core/src/codex.rs

@@ -4904,6 +4904,8 @@ mod handlers {
     use crate::config_loader::CloudRequirementsLoader;
     use crate::config_loader::LoaderOverrides;
     use crate::config_loader::load_config_layers_state;
+    use crate::realtime_context::REALTIME_TURN_TOKEN_BUDGET;
+    use crate::realtime_context::truncate_realtime_text_to_token_budget;
     use codex_features::Feature;
     use codex_utils_absolute_path::AbsolutePathBuf;
 
@@ -5124,6 +5126,10 @@ mod handlers {
         if text.is_empty() {
             return;
         }
+        let text = truncate_realtime_text_to_token_budget(&text, REALTIME_TURN_TOKEN_BUDGET);
+        if text.is_empty() {
+            return;
+        }
         if sess.conversation.running_state().await.is_none() {
             return;
         }