Per-branch Terminology cache location could cause issues with CI builds

[dotasek]

When the Publisher finds the environment variables set by Azure CI builds, it automatically creates a Terminology cache directory at ~/.fhir/{ghOrg}/{ghRepo}/{ghBranch}.

This results in two issues:

1. This directory will persist and any contents will be present if the PR is rebuilt. If the cache contains any invalid responses, the next build on that PR could pick those up.

2. The cache is located on the agent machine, NOT in the generated agent work directory. If the first build of the PR happened on GCP1, then only GCP1 will have the cached responses. If the next build runs on GCP2, the cache will be regenerated on GCP2 and will not re-use the contents of GCP1.

According to @grahamegrieve, the intention of this location was to re-use the cache. This should speed up subsequent builds. The cost of this is the possibility of issue 1 occurring, but the tradeoff has been deemed to be worth it.

However, the advantage of cache re-use is unavailable now that we have multiple agent machines, since issue 2 is very much occurring.

A location re-usable across builds would be ideal.