Prevent postgres deadlocking

[clarafu]

When we update the cache indexes for the all the pipelines that use a
resource config which we found new versions for, postgres can get into a
deadlock when two resource configs try to update the same pipeline's
cache indexes at the same time. By running all the cache index updates
by itself (not in a transaction), we will prevent that from happening.

[vito]

note: there's probably a better way to do this but I think it can wait until we start on bigger architecture changes.

this does the same amount of updates pre-global-resources, and there will still be fewer queries overall.