tests: add MoE subgraph tests with require_full_compilation enforcement

Add tests/py/dynamo/hlo/test_moe.py covering all 7 MoE routing and
dispatch variants from popular open-source LLMs (Mixtral, Qwen2, Qwen3,
Llama4, DeepSeek-V2, DeepSeek-V3/R1, Nemotron-H).  50 parameterized
test cases verify TRT numerical correctness against PyTorch reference.

Extend DispatchTestCase.run_test() with require_full_compilation=True
support: calls TRTInterpreter.validate_conversion() before building the
engine and fails immediately if any op lacks a TRT converter.  All MoE
tests pass this check, confirming zero PyTorch fallback for every
routing/dispatch/expert-MLP combination tested.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: yizhuoz004

tests/py/dynamo/conversion/harness.py

@@ -422,6 +422,7 @@ def run_test(
         immutable_weights=True,
         use_explicit_typing=False,
         decompose_attention=False,
+        require_full_compilation=False,
     ):
         # TODO: lan to remove this and set use_dynamo_traccer to True by default
         # once all the converter test files are moved to use_dynamo_tracer
@@ -434,6 +435,7 @@ def run_test(
             immutable_weights=immutable_weights,
             use_explicit_typing=use_explicit_typing,
             decompose_attention=decompose_attention,
+            require_full_compilation=require_full_compilation,
         )
 
         mod = self.generate_graph(
@@ -499,6 +501,13 @@ def run_test(
             compilation_settings=compilation_settings,
         )
 
+        if require_full_compilation:
+            missing = interp.validate_conversion()
+            self.assertFalse(
+                missing,
+                f"require_full_compilation=True but the following ops have no TRT converter: {missing}",
+            )
+
         super().run_test(
             mod,
             trt_inputs,