Example 64 (Ada FP8 grouped GEMM) refuses capable GPUs via exact CC 8.9 gate

Description

examples/64_ada_fp8_gemm_grouped/ada_fp8_gemm_grouped.cu (~line 1107) gates execution on exactly compute capability 8.9:

if (!(properties.major == 8 && properties.minor == 9)) {
    std::cerr << "... requires a device of compute capability 89." ...

The grouped FP8 kernels here compile cleanly for newer architectures, and the sibling example 58_ada_fp8_gemm uses the inclusive form:

if (properties.major < 8 || (properties.major == 8 && properties.minor < 9))

On an RTX 5060 Ti (sm_120a, CUDA 12.8.1) the stock binary exits immediately with the capability message; with the gate relaxed to the sibling's form the example builds and runs to completion including device-reference verification (Passed). Example 64 is therefore refusing capable hardware for no functional reason.

Suggested fix

Use the same major*10+minor < 89 / major < 8 || (major == 8 && minor < 9) inclusive check as example 58.

Open examples/64_ada_fp8_gemm_grouped/ada_fp8_gemm_grouped.cu around line 1107 and compare its compute-capability check with the inclusive check in example 58_ada_fp8_gemm. Update the gate to accept architectures at or above the stated minimum, then build and run Example 64 to confirm device-reference verification reports “Passed” on newer capable hardware.