Recover capability discovery after executor disconnects (#38420)

## Why

Transient executor disconnects could leave capability discovery and skill catalogs
stuck on a cached failure for the rest of a thread, even after the executor
reconnected.

## What changed

- Replay capability discovery after executor recovery and retry transient failures
  on later requests while continuing to cache permanent failures.
- Avoid caching skill catalogs produced from failed discovery so a later step can
  load the recovered catalog.
- Mark the MCP runtime dirty when recovered manifests change the projected MCP
  servers, and allow discovery to be cancelled with the turn.

## Testing

- Cover same-request recovery after a disconnect and recovery on a later request.
- Cover retry classification through connection-attempt errors and skill catalog
  caching after discovery recovers.

GitOrigin-RevId: a57f90844351e73ea831931f72a9ddc4e4f3335c

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

@@ -422,11 +422,14 @@ impl Session {
             })
             .cloned()
             .collect::<Vec<_>>();
-        Some(Arc::new(
-            cache
-                .snapshot(&selected_capability_roots, &sandbox_contexts)
-                .await,
-        ))
+        let discovery = cache
+            .snapshot(&selected_capability_roots, &sandbox_contexts)
+            .await;
+        if cache.take_recovered_discovery() {
+            // Root selection is unchanged, but recovered manifests can change MCP servers.
+            self.mark_mcp_runtime_dirty();
+        }
+        Some(Arc::new(discovery))
     }
 
     pub(crate) async fn resolve_selected_capability_roots_for_step(