Add unit tests for the FP8 variant of expert-parallel MoE dispatch.
Source: look for ep_moe_expert_dispatch_fp8 in custom_ops/gpu_ops/
Registration: custom_ops/gpu_ops/cpp_extensions.cc
Test file: tests/operators/test_ep_moe_expert_dispatch_fp8.py
Should test token routing to experts with FP8 quantization. Compare dispatch results against a pure Python reference.
Branch: task/048-ep-moe-dispatch-fp8-test