opencl: ref count `ggml_backend_opencl_context` and refactor profiling (llama/14254)

opencl: add `mul_mv_id_q4_0_f32_8x_flat` (llama/14003)

opencl: add `backend_synchronize` (llama/13939)

OpenCL: Add concat, tsembd, upscale, tanh, pad and repeat (llama/13840)

opencl: add new ops - `argsort`, `div`, `sub`, `addrows`, `sigmoid`, `group_norm` (llama/13787)

opencl: mark `mul_mat` `f32f32` as supporting non-contiguous tensors (llama/13790)

opencl: Add support for multiple devices (llama/12622)

opencl: fix couple crashes (llama/12795)

opencl: remove unnecessary assert for `add` (llama/13257)

opencl: split ggml-opencl.cl into multiple files and cleanup (llama/12886)

opencl: fix incorrect local_size index in profiling log (llama/12868)

opencl: better identify Adreno GPU (llama/12760)

opencl: use `max_alloc_size` in backend ctx instead of querying again (llama/12705)

opencl : fix memory allocation size (llama/12649)

opencl: add multi and vision rope, `gelu_quick` and `im2col` (llama/12600)

opencl: improve profiling (llama/12442)

opencl: use OpenCL C standard supported by the device (llama/12221)

opencl: Noncontiguous `norm`, `rms_norm`, disable `fp16` for some ops (llama/12217)

opencl : fix buffer alignment (llama/12197)

opencl : fix `ulong` kernel args were set from `int` variables (llama/12174)

opencl : fix profile-related errors (llama/12095)

ggml : upgrade init_tensor API to return a ggml_status (llama/11854)

opencl: fix for small models (llama/11950)

opencl: Fix rope and softmax (llama/11833)

ggml : add opencl backend (skip) (llama/10693)