Add inductor cache metrics (pytorch#138603)

Each inductor event should have exactly one hit, miss, bypass etc. Add it to the inductor compile event.

Add triton_compile as a compiler phase with `dynamo_timed`. This way, we get PT2 Compile Event Logs for triton as well.

Here's what triton events look like:  {F1941513932}
And this on a cache hit(since we still redo this work):
 {F1941514350}

Inductor cache info:
 {F1941528530}

Differential Revision: [D64703392](https://our.internmc.facebook.com/intern/diff/D64703392/)

@diff-train-skip-merge

Pull Request resolved: pytorch#138603
Approved by: https://github.com/oulgen

Author: jamesjwu

torch/_inductor/codecache.py

@@ -1512,6 +1512,18 @@ def load(  # type: ignore[no-untyped-def]
             cache_info["cache_event_time"],
             metadata=cache_info,
         )
+        # Add event data about cache hits/miss
+        # TODO: add remote cache get/put timings here too
+        chromium_log.add_event_data(
+            "inductor_compile",
+            cache_state=cache_state,
+            cache_event_time=cache_info["cache_event_time"],
+            key=cache_info.get("key"),
+            components=cache_info.get("components"),
+            cache_bypass_reason=cache_info.get("cache_bypass_reason"),
+            remote_cache_enabled=remote,
+            local_cache_enabled=local,
+        )
         torch._logging.trace_structured(
             "artifact",
             metadata_fn=lambda: {