Add TRT plugin custom ops support

[chilo-ms]

Summary

Register TensorRT plugins as custom ops under the trt.plugins domain, enabling ORT to recognize and route nodes with TRT plugin op types to the TensorRT EP.

Reference: upstream ORT implementation

Changes

New files

• src/tensorrt_execution_provider_custom_ops.h - Defines TensorRTCustomOp (using Ort::CustomOpBase) and TensorRTCustomKernel structs with variadic inputs/outputs, since TRT plugin API doesn't expose I/O counts.
• src/tensorrt_execution_provider_custom_ops.cc - Implements CreateTensorRTCustomOpDomainList():
  • Loads extra TRT plugin libraries (LoadLibraryA/dlopen)
  • Initializes default TRT plugins via initLibNvInferPlugins
  • Enumerates all registered plugin creators (V1 and V3One types)
  • Creates custom ops under the trt.plugins domain
  • Uses static storage with mutex for thread-safe, process-lifetime persistence

Modified files

• src/tensorrt_provider_factory.h - Added GetNumCustomOpDomainsImpl/GetCustomOpDomainsImpl static methods and custom_op_domain_list_ member.
• src/tensorrt_provider_factory.cc - Wired the two function pointers in the constructor and implemented the static methods.

Tests

• tests/tensorrt_basic_test.cc - Added TRTPluginsCustomOpTest adapted from the upstream TensorrtExecutionProviderTest.TRTPluginsCustomOpTest. Validates that a model with trt.plugins domain ops can be loaded and session creation succeeds.
• tests/testdata/trt_plugin_custom_op_test.py - Python script to generate the test model.
• tests/testdata/trt_plugin_custom_op_test.onnx - Generated test model with DisentangledAttention_TRT node.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}