Handle concurrent vhost creation and deletion better

In the case of a vhost creation and immediate deletion those
operations can end up executing concurrently. This changes
a number of things to be more resilient:

 * Default vhost restart strategy changes to "continue" (previosly "ignore").
   It's a tough default to pick but we believe the damage of a single vhost terminating
   the entire node that may have a lot of users is greater than the need to set up an alternative
   strategy for environments where only one vhost is used. Note that the event of
   message store or vhost supervision tree termination should be very rare.

   Per discussion with @kjnilsson.

 * For vhost [message store] recovery, we now log "no_such_vhost" errors
   in a sensible way instead of a bunch of process terminations.

 * Max restart intensity settings for rabbit_vhost_sup_wrapper are
   changed to accommodate integration test suites that rapidly set up
   temporary vhost and delete them. In most failure message store scenarios
   the new value is still low enough to lead to vhost supervision tree
   termination when recovery is impossible.

There are also drive by doc changes.

Fixes #1280.

0 files changed, 0 insertions, 0 deletions