Add documentation on how to extend base prepackaged servers with new images (xgboost, sklearn, etc)

[axsaucedo]

Currently we get questions such as #2059 that ask for newer versions of sklearn. Given the complexity to support multiple versions of multiple ML libraries on every release is too complex, we should improve documentation to show basically how people can build their own prepackaged servers and override the base images in the configmap. We currently already have documentation on how to create custom wrappers via https://docs.seldon.io/projects/seldon-core/en/latest/servers/custom.html so it may just be a matter of adding a few jupyter notebooks to show how someone can actually just re-build the base images with their custom requirements and overriding the configmap.

This is related to the parent task of extending documentation on more advanced components of seldon core #2012