Use Metalhead.jl for the vision examples

Currently, the examples for [ConvMixer](https://github.com/FluxML/model-zoo/blob/dce9dd0d44567ee9dabccba9b92b931298b462ce/vision/convmixer_cifar10/convmixer.jl#L6) and [VGG](https://github.com/FluxML/model-zoo/blob/dce9dd0d44567ee9dabccba9b92b931298b462ce/vision/vgg_cifar10/vgg_cifar10.jl#L39) have the code for the models written inside them. Since Metalhead has cleaner versions of these models, would it make sense to replace these or are these code examples kept specifically?

Compare the inline model definitions in vision/convmixer_cifar10/convmixer.jl and vision/vgg_cifar10/vgg_cifar10.jl with the corresponding Metalhead models. First determine whether these examples are intended to teach model construction or demonstrate complete models; done requires a documented decision and consistent updates to both examples if reuse is chosen.

Written by the indexing model from the issue text.