Prevent streaming volumes between concourse workers.

[kallisti5]

We now have two concourse workers, however they waste infinite time attempting to stream build step artifacts back and forth between them.

Concourse seems to assume workers all exist in the same network... which is definitely not the case for us :-1:

Here's a screenshot of it wasting over 20 minutes copying the git artifacts from the first node to the second node...

https://github.com/concourse/concourse/issues/8861 suggests we have to target pipelines to tags on specific workers :disappointed: That essentially means we have the following options:

• Rigidly assign pipelines and resources to individual nodes (removing the fault tolerance ability)
• Play with placement strategies.. (though, the default should be our "ideal") https://concourse-ci.org/container-placement.html#fewest-build-containers-strategy
• Figure out a way for artifact streaming to be quick between geographically diverse nodes.

[kallisti5]

in the case above... hbrl01 was "stalled"

name    containers  platform  tags  team  state    version  age
hbav01  27          linux     none  none  running  2.5      18d

the following workers have not checked in recently:
name    containers  platform  tags  team  state    version  age
hbrl01  18          linux     none  none  stalled  2.5      30d

So not even sure what's going on there. I guess it "stalled" while receiving the data from mine?