Prerequisites

• I am running the latest code. Mention the version if possible as well.
• I carefully followed the README.md.
• I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• I reviewed the Discussions, and have a new and useful enhancement to share.

Feature Description

Currently various GGML backends have their own logging routines. These need to be unified and updated to behave similar to those defined in llama-impl.h/llama.h.

Motivation

Cleans up redundant logging behavior.

Possible Implementation

This change will minimally require:

1. Updating ggml-impl.h with a series of GGML_LOG_* macros for the various log levels.
2. Adding a ggml_log_set routine to ggml.h.
3. Adding internal/default logging callbacks in ggml.cpp/ggml-impl.h.
4. Modifying the llama_log_set routine to internally call ggml_log_set so they use the same logging mechanism.
5. Remove backend-specific log-setup logic (e.g. ggml_backend_metal_log_set_callback &co.), as registered in the current llama_log_set routine.
6. For each backend, replace back-end specific log invocations (e.g. VK_LOG_DEBUG) with the new unified macros.