Add shadow metrics for lexical skill selection (#32761)

## What changed

- Add an opt-in `skill_search` feature that ranks prompt-visible skills against each turn's user input with a bounded weighted lexical selector.
- Keep the ranked selection out of model-visible context and record metrics for selection cost, catalog reduction, and whether later implicit or `skills.read` invocations matched the ranked candidates.
- Include host-provided skills in the experiment catalog without changing the rendered skill catalog.

## Testing

- Add selector unit tests covering ranking, limits, truncation, stop words, and deterministic tie-breaking.
- Add extension tests covering turn-local invocation recording and host-skill shadow selection.

GitOrigin-RevId: 4d00a1c805ea8b391d6c6ac6a8450afa88ca3e25

codex-rs/app-server/src/extensions.rs

@@ -84,13 +84,17 @@ where
         .with_orchestrator_provider(Arc::new(
             codex_skills_extension::OrchestratorSkillProvider::new(),
         ));
-    codex_skills_extension::install_with_providers(
+    codex_skills_extension::install_with_providers_and_metrics(
         &mut builder,
         skill_providers,
+        codex_otel::global(),
         |config: &Config| codex_skills_extension::SkillsExtensionConfig {
             include_instructions: config.include_skill_instructions,
             bundled_skills_enabled: config.bundled_skills_enabled(),
             orchestrator_skills_enabled: config.orchestrator_skills_enabled,
+            shadow_selection_enabled: config
+                .features
+                .enabled(codex_features::Feature::SkillSearch),
         },
     );
     Arc::new(builder.build())