sycl: Add more debug prints

[Rbiessy]

This adds more debugging prints related to the operations being run. scope_op_debug_print is introduced to print the "call ..." and "call ... done" logs in a single line.
This adds more information for all the operations and its destination and input tensors: the tensor type, its number of elements, stride sizes and whether the tensor is strided (i.e. non-contiguous) or permuted.
For mul_mat this also add some debug prints when conversations kernels are called from and to fp16, fp32 and quantized types.
This also adds debugging prints for a few operations that were missing it.

The output for llama-2-7b.Q4_0 PP and TG is added here as an example:

• llama-2-7b.Q4_0__log_bench_pp_512_no_warmup.txt
• llama-2-7b.Q4_0__log_bench_tg_1_no_warmup.txt

[NeoZhangJianyu]

I have a concern that this dynamic method will decrease the performance of running time:

The legacy method won't impact the performance when disable the debug mode.
But this method will spend time to create the object when disable the debug mode.

Have you test the performance impact with legacy code when disable the debug mode?

[qnixsynapse]

There is a check here in the function:

if (!g_ggml_sycl_debug) {
        return;
    }

So, if debugging mode is disabled, it will not do anything so no performance hit in non debug mode I think.

[NeoZhangJianyu]

scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/1);

this code will trigger constructor function when entry ggml_sycl_dup().
When call ggml_sycl_dup() n times, the constructor function will be called n times.

Legacy code is the static code. It won't take more time when entry ggml_sycl_dup().

[qnixsynapse]

Nice.
Is it possible to extend this to backend interface functions as well such as init, set, memset, clear, reset, etc?