* ggml : add BF16 support to ROPE and SET_ROWS (CPU, CUDA) * cuda : SET_ROWS from BF16 into quantized types and F8_E4M3 * cuda : fuse BF16 rope into a BF16 cache * tests : trim the BF16 rope cases to one per code path
* ggml : add BF16 support to ROPE and SET_ROWS (CPU, CUDA) * cuda : SET_ROWS from BF16 into quantized types and F8_E4M3 * cuda : fuse BF16 rope into a BF16 cache * tests : trim the BF16 rope cases to one per code path