layer_norm for 2D and more?

Hi all,

Is it useful to have a **layer_norm_2d** (the input is a matrix) in addition to layer_norm?
Currently, I tried a naive version but it may be slow.

```
Expression layer_norm_2d(const Expression& x, const Expression& g, const Expression& b){
std::vector vCols(x.dim().d[1]);
for (unsigned i = 0; i < x.dim().d[1]; i++){
Expression c_x = select_cols(x, {i});
vCols[i] = layer_norm(c_x, g, b);
}
return concatenate_cols(vCols);
}
```
Any suggestion? I would be happy to work on this!

Thanks!

Start by reading the existing layer_norm entry point and the related select_cols and concatenate_cols operations shown in the issue. Determine the intended normalization dimensions and parameter shapes for matrix input, then compare the result and performance with the proposed column-wise approach. Done means the behavior and API are agreed on and covered by relevant tests.

Written by the indexing model from the issue text.