Caching of downloaded function dependencies

[ksatzke]

Currently, function dependencies which are configured under "Requirements" in the GUI are downloaded from scratch whenever this function or the corresponding workflow is deployed.

However, depending on the involved dependency packages, in some cases (tensorflow, torch, ...) these downloads can become quite large. Therefore, it would be advantageous to cache already downloaded packages, so that they can be used by other deployments and need to be downloaded only once.