BUG/MEDIUM: peers: don't fail twice to grab the update lock

BUG/MEDIUM: peers: don't fail twice to grab the update lock

When the expire task is running fast (i.e. running almost alone), it's
super hard to grab the update lock and peers can easily trigger the
watchdog because the time it takes to grab this lock is multiplied by
the number of updates to perform. This is easier to trigger at the end
of an injection session where the expire task is omni-present. Let's
just record that we failed once and don't fail a second time in the
loop.

This should be backported to 3.2, but probably not further given that
this area changed significantly in 3.2.