Fix LayerNorm gradient flow issue

[tymat]

• Fix LayerNorm.forward() to use tensor operations instead of scalar operations
• Replace sum_keepdim()/size with mean_keepdim() to preserve gradients
• Use broadcast_add() with epsilon tensor instead of scalar addition
• Fix ops::layer_norm_slow() with same gradient-preserving changes
• Update ops::layer_norm() to use slow implementation for proper gradients
• Add comprehensive gradient flow test (now passes with 100% gradient flow)
• Add numerical equivalence test to ensure accuracy is maintained
• Fixes training issues where LayerNorm parameters weren't being updated

Resolves gradient propagation bug where only 33% of parameters received gradients during backpropagation, preventing proper model training. #3011

[AlpineVibrations]

great. it would be awesome to have more training code examples and workflows with candle

[ivarflakstad]

Hey! Thanks for this :)

I think we'll have to implement this in the optimized kernels as well before we can merge. I assume all the variants (cpu, cuda, metal) suffer from the same issue?