Add spans to run_turn (#27107)

## Why
Codex app-server latency traces do not granularly cover turn
orchestration, sampling-request preparation, and tool-loading work.
These spans help separate local coordination/setup costs from model
streaming and tool execution.

## What changed
- Add `run_turn.*` spans around sampling-request input preparation and
post-sampling state collection
- Add function-level trace spans around turn setup, hook execution,
compaction, prompt construction, and MCP tool exposure
- Add `built_tools.*` spans around plugin loading and discoverable-tool
loading

## Verification
Trigger Codex rollout and observe new spans are included

codex-rs/core/src/hook_runtime.rs

@@ -36,6 +36,7 @@ use codex_protocol::protocol::SessionSource;
 use codex_protocol::protocol::SubAgentSource;
 use codex_thread_store::ReadThreadParams;
 use serde_json::Value;
+use tracing::instrument;
 
 use crate::context::ContextualUserFragment;
 use crate::context::HookAdditionalContext;
@@ -97,6 +98,7 @@ impl From<UserPromptSubmitOutcome> for ContextInjectingHookOutcome {
     }
 }
 
+#[instrument(level = "trace", skip_all)]
 pub(crate) async fn run_pending_session_start_hooks(
     sess: &Arc<Session>,
     turn_context: &Arc<TurnContext>,
@@ -291,6 +293,7 @@ pub(crate) async fn run_post_tool_use_hooks(
     outcome
 }
 
+#[instrument(level = "trace", skip_all)]
 pub(crate) async fn run_turn_stop_hooks(
     sess: &Arc<Session>,
     turn_context: &Arc<TurnContext>,
@@ -427,6 +430,7 @@ pub(crate) async fn run_post_compact_hooks(
     }
 }
 
+#[instrument(level = "trace", skip_all)]
 pub(crate) async fn run_legacy_after_agent_hook(
     sess: &Arc<Session>,
     turn_context: &Arc<TurnContext>,