Add max context window model metadata (#18382)

Adds max_context_window to model metadata and routes core context-window
reads through resolved model info. Config model_context_window overrides
are clamped to max_context_window when present; without an override, the
model context_window is used.

codex-rs/core/src/memories/prompts.rs

@@ -239,7 +239,7 @@ pub(super) fn build_stage_one_input_message(
     rollout_contents: &str,
 ) -> anyhow::Result<String> {
     let rollout_token_limit = model_info
-        .context_window
+        .resolved_context_window()
         .and_then(|limit| (limit > 0).then_some(limit))
         .map(|limit| limit.saturating_mul(model_info.effective_context_window_percent) / 100)
         .map(|limit| (limit.saturating_mul(phase_one::CONTEXT_WINDOW_PERCENT) / 100).max(1))

codex-rs/core/src/memories/prompts_tests.rs

@@ -41,6 +41,7 @@ fn build_stage_one_input_message_uses_default_limit_when_model_context_window_mi
     let input = format!("{}{}{}", "a".repeat(700_000), "middle", "z".repeat(700_000));
     let mut model_info = model_info_from_slug("gpt-5.2-codex");
     model_info.context_window = None;
+    model_info.max_context_window = None;
     let expected_truncated = truncate_text(
         &input,
         TruncationPolicy::Tokens(phase_one::DEFAULT_STAGE_ONE_ROLLOUT_TOKEN_LIMIT),

codex-rs/core/src/session/turn_context.rs

@@ -75,9 +75,11 @@ pub(crate) struct TurnContext {
 impl TurnContext {
     pub(crate) fn model_context_window(&self) -> Option<i64> {
         let effective_context_window_percent = self.model_info.effective_context_window_percent;
-        self.model_info.context_window.map(|context_window| {
-            context_window.saturating_mul(effective_context_window_percent) / 100
-        })
+        self.model_info
+            .resolved_context_window()
+            .map(|context_window| {
+                context_window.saturating_mul(effective_context_window_percent) / 100
+            })
     }
 
     pub(crate) fn apps_enabled(&self) -> bool {