Missing neural network operations

## NNlib

- [x] activations
- [x] tanh_fast
- [x] sigmoid
- [x] sigmoid_fast
- [x] softmax!
- [x] logsoftmax!
- [x] conv / conv! (#258)
- [x] maxpool / maxpool! (#258)
- [x] meanpool / meanpool! (#258)
- [x] batched_mul / batched_mul! (#258)
- [x] padding
- [x] pad_constant
- [x] pad_circular
- [x] pad_repeat
- [x] pad_zeros
- [x] pad_reflect
- [x] pad_symmetric
- [x] ∇conv_data / ∇conv_data!
- [x] ∇conv_filter / ∇conv_filter!
- [x] gather / gather!
- [x] scatter / scatter! https://github.com/EnzymeAD/Reactant.jl/pull/1395
- [x] ~dot_product_attention~
- [x] ~pixel_shuffle~
- [x] batchnorm (has a special stablehlo impl)
- this needs to wait for a corresponding adjoint on EnzymeJAX end
- [ ] grid_sample / grid_sample!
- [x] upsampling
- [x] ~nearest~
- [x] linear
- [x] bilinear
- [x] trilinear
- [x] fold / fold!
- [x] unfold / unfold!

Some of these might not even be needed, but we should test the generated IR. Strikethrough denotes we don't need specialized handling for these operations

Feel free to add missing operations to the list