Add SME Depthwise Convolution Planar Kernels and Reference implementation.

- Add example use of planar kernels with padding.
- Planar kernels produce four output rows given an input pointer with stride and padding arguments
- Adds packing function for weights packing with depthwise kernels.
- Packing function packs based on vector length along channel dim.

Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>