Expose profiler advanced configuration as a Python dict.

In profiler.cc, the advanced_configuration property of tensorflow::ProfileOptions is now exposed as a Python dictionary. The getter converts the proto map to a nb::dict, handling different value types (bool, int64, string).

Example error:
```
ProfileOptions().advanced_configuration
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: Unable to convert function return value to a Python type! The signature was
    (self) -> proto2::Map<std::__u::basic_string<char, std::__u::char_traits<char>, std::__u::allocator<char>>, tensorflow::ProfileOptions_AdvancedConfigValue>
```
PiperOrigin-RevId: 841944949

Author: lukebaumann

tests/profiler_test.py

@@ -30,6 +30,7 @@
 import jax.numpy as jnp
 import jax.profiler
 import jax._src.test_util as jtu
+import jaxlib
 
 from jax._src import profiler
 from jax import jit
@@ -508,5 +509,18 @@ def on_profile():
         unittest.mock.ANY,
     )
 
+  @unittest.skipIf(jaxlib.version <= (0, 8, 2), "advanced_configuration getter is added in jaxlib 0.8.2")
+  def test_advanced_configuration_getter(self):
+    options = jax.profiler.ProfileOptions()
+    advanced_config = {
+        "tpu_trace_mode": "TRACE_COMPUTE",
+        "tpu_num_sparse_cores_to_trace": 1,
+        "enableFwThrottleEvent": True,
+    }
+    options.advanced_configuration = advanced_config
+    returned_config = options.advanced_configuration
+    self.assertIsInstance(returned_config, dict)
+    self.assertEqual(returned_config, advanced_config)
+
 if __name__ == "__main__":
   absltest.main(testLoader=jtu.JaxTestLoader())