Adding VGG that works for neon v2.3 with MKL backend

[wei-v-wang]

training and inference works fine with MKL GPU training occasionally work GPU inference works fine.

Include updated Readme regarding new format of weights.