This was the main annoyance with the existing script, it would fetch
one repository after the other instead of multiple in parallel.
The `xe` tool is a simple `xargs`-like that can do that.
We put the json information for all repositories into a file with one
object per line, then pass that to xe.