Make opam shared root initialization more robust

This:

• drops the previous version of the tooling
• installs the new tooling:
  • use set -e in the main script so command failures are immediately detected by icinga
  • the script triggered by the timer/service script updates all instances at once (in working order, first opam root with the default opam instance, then the other secondary repositories). Implementation wise, it's looping over all instances to init or update them all.
• also changes the timer period to ~24h (this matches the scheduler configuration for

the opam lister)

Related to T3826