[TOPI][NN][Layer_Norm] Fix layer_norm error with reduce-only axes

[eaten-cake]

Fixes #18002
When all dimensions of layer_norm input are reduction axes, the variable non_reduce_indices becomes empty, while the previous variable target_shape has atleast1d parameter enabled, causing shape inconsistency.

tvm/include/tvm/topi/nn/layer_norm.h

Lines 111 to 121 in 2d964b4

	auto layer_norm_func = [&](const Array<Var>& indices) {
	Array<Var> reduce_indices, non_reduce_indices;
	for (int i = 0, n = static_cast<int>(indices.size()); i < n; ++i) {
	if (std::find(real_axis.begin(), real_axis.end(), i) != real_axis.end()) {
	reduce_indices.push_back(indices[i]);
	} else {
	non_reduce_indices.push_back(indices[i]);
	}
	}
	auto mean = temp_x(non_reduce_indices) / reduce_extent;
	auto var = temp_x2(non_reduce_indices) / reduce_extent - mean * mean;

tvm/include/tvm/topi/nn/layer_norm.h

Lines 67 to 68 in 2d964b4

	auto target_shape =
	MakeReduceTargetShape(real_axis, data, /*keepdims=*/false, /*atleast1d=*/true);

Also added tests for reduce-only axes cases

[eaten-cake]

cc @MasterJH5574 @Hzfengsy @tqchen @mshr-h