ggml : relax scale to non-contiguous views (#31146) * ggml : relax scale to non-contiguous views SCALE requires src0 to be 1D-contiguous. Relax this to dim-0 contiguity with arbitrary row strides (ggml_is_contiguous_rows), same as the unary ops in the Metal backend. - ggml: relax the assert in ggml_scale_impl - cpu: address src0 rows through their byte strides (f32, bf16) - cuda: stride-aware kernel, fastdiv row decomposition - metal: support bf16 scale (guarded unary instantiations) - sycl: keep the gate at contiguous, the kernel is flat f32 Add view variants to test_scale. Verified on CPU, Metal and CUDA. Assisted-by: dsh:llama.cpp/ggml-org/Qwen3.8-Flash-Next-GGUF (RTX 5090 32GB, DDR5 128GB) * cuda : skip the softcap fusion on non-contiguous views The fused softcap kernel reads src and w
ggml：放宽对非连续视图的缩放（#31146）* ggml：放宽对非连续视图的缩放 SCALE 要求 src0 必须是一维连续内存。现将其放宽为在维度 0 上具有任意行步长的连续性（ggml_is_contiguous_rows），这与 Metal 后端中的单目运算保持一致。 - ggml：放宽 ggml_scale_impl 中的断言 - cpu：通过字节步长访问 src0 的各行（f32、bf16） - cuda：支持步长感知的内核，采用 fastdiv 行分解 - metal：支持 bf16 缩放（通过受保护的单目实例化） - sycl：保持 gate 处于连续状态，内核则以 f32 平铺形式实现 为 test_scale 添加视图变体。已在 CPU、Metal 和 CUDA 上验证。协助者：dsh:llama.cpp/ggml-org/Qwen3.8-Flash-Next-GGUF（RTX 5090 32GB，DDR5 128GB）* cuda：在非连续视图上跳过软截断融合 融合后的软截断内核会读取 src 和 w
