Allow more flexible article cache creation and handling

Currently there is an artifact cache that is created on-demand based on artifact retrieval requests where agents request downloads of entire folders.

What seems to happen is that

• zip directory artifact request comes in
• background thread is created to zip the artifacts
• agent/client gets a 202 accepted
• zip is stored in godata/artifacts/cache with structure artifacts/pipelines/...
• agent/client needs to retry. There is a sort of random back-off logic, but sometimes creation takes too long and it will timeout
  • wait 10-20s, retry 1
  • wait 20-30s, retry 2
  • wait 30-40s, retry 3
  • fail (after 60-90s, mean 75s)

The challenges are

• this cache folder is not documented anywhere
• generally it is inside the godata area and so is on persistent storage
• there is no normal or periodic cleanup process for it
• there is no way to change its location to be outside the normal artifactsDir defined for the server.
• use of zips means executable bits and metadata are lost, however may have other benefits
• creation of zips is CPU limited to a single thread/core