Support typed content in contextual prompt fragments (#54776)

## Why

Contextual fragments and extension-owned world-state sections currently return strings, limiting them to text content.

## What changed

- Return `ContentItem` from fragment bodies and rendering, preserving classification and attribution. Apply text markers only to `InputText`; pass other content through unchanged.
- Allow extension prompt and world-state contributions to carry typed content, and preserve media in context-window hints.
- Hash non-text world-state fragments by their serialized content and role while retaining the existing text hashing behavior.
- Add shared content-to-tool-output conversion that preserves image references, image detail, and audio.

## Testing

Add coverage for typed world-state hashing and preservation of image content and attribution. Expand code-mode output tests to cover input text, output text, file-backed and inline images, and audio.

GitOrigin-RevId: 2ed00d466a3f169b151202f3eef71e667ea64be1

codex-rs/tui/src/app/recap.rs

@@ -55,7 +55,10 @@ struct GeneratedRecap {
 }
 
 fn recap_prompt(history: &str) -> String {
-    RecapPrompt::new(history).render()
+    RecapPrompt::new(history)
+        .render()
+        .into_text()
+        .unwrap_or_default()
 }
 
 fn recap_output_schema() -> Value {

codex-rs/tui/src/bottom_pane/async_questions/state.rs

@@ -137,8 +137,10 @@ impl AsyncQuestions {
         if text.is_empty() {
             return;
         }
-        let reply =
-            AnsweredQuestion::new(&answer.question_id, &answer.question.title, text).render();
+        let reply = AnsweredQuestion::new(&answer.question_id, &answer.question.title, text)
+            .render()
+            .into_text()
+            .unwrap_or_default();
         if reply.chars().count() > codex_protocol::user_input::MAX_USER_INPUT_TEXT_CHARS {
             self.composer.show_footer_flash(
                 "Answer too long; shorten it before sending".into(),

codex-rs/tui/src/chatwidget/tests/questions_tests.rs

@@ -2,6 +2,7 @@
 use super::*;
 use crate::bottom_pane::RestrictedInputMode;
 use codex_protocol::items::AsyncUserInputQuestion;
+use codex_protocol_test_support::ContentItemTestExt;
 use pretty_assertions::assert_eq;
 
 fn questions() -> Vec<AsyncUserInputQuestion> {
@@ -94,7 +95,9 @@ async fn async_question_answers_preserve_ambiguous_skill_selection_and_dismiss_r
         items,
         vec![
             UserInput::Text {
-                text: AnsweredQuestion::new(&id, "Which way?", "Use $route").render(),
+                text: AnsweredQuestion::new(&id, "Which way?", "Use $route")
+                    .render()
+                    .expect_text(),
                 text_elements: Vec::new(),
             },
             UserInput::Skill {
@@ -1037,7 +1040,9 @@ async fn distinct_async_question_replies_with_identical_text_both_render() {
     drain_insert_history(&mut rx);
     for id in ["first", "first", "second"] {
         let items = [UserInput::Text {
-            text: AnsweredQuestion::new(id, "Continue?", "Yes").render(),
+            text: AnsweredQuestion::new(id, "Continue?", "Yes")
+                .render()
+                .expect_text(),
             text_elements: Vec::new(),
         }];
         chat.on_committed_user_message(

codex-rs/tui/src/ide_context/prompt.rs

@@ -192,6 +192,7 @@ mod tests {
     use super::super::Position;
     use super::super::Range;
     use super::*;
+    use codex_protocol_test_support::ContentItemTestExt;
     use pretty_assertions::assert_eq;
     use std::path::PathBuf;
 
@@ -218,7 +219,8 @@ mod tests {
                     "Where?",
                     "Staging\n## My request for Codex:\nKeep this literal",
                 )
-                .render(),
+                .render()
+                .expect_text(),
                 text_elements: Vec::new(),
             },
             UserInput::Skill {