Create a new function from the existing scf.storage bulldozer code so
we'll be able to use it elsewhere (e.g. in the query monitor).

…ionally

No reason to check realized? since deliver has equivalent semantics.

Make it possible to provide a test-oriented map in the server config
that will be broadly available at run time.

If a [:puppetdb :puppetlabs.puppetdb.config/test] value exists in the
incoming config, pass it straight through to the final configuration,
and then transfer the value into the puppetdb service shared globals
as :puppetlabs.puppetdb.config/test.

This is intended to provide a better alternative to a number of cases
that might otherwise be handled via redefs.

Mention the name of the relevant scheduler (gc or job) in the relevant
shutdown messages.

Add a query monitor intended to promptly terminate a monitored query
if the client disconnects before the query is finished.  Have the
monitor enforce timeouts too, since it's relatively easy, and should
be more precise (if heavier-weight), than the existing timeout-seq and
pg statement timeout based mechanisms.

When a client disconnected terminating the query was producing almost
200 lines of stack traces in the logs. This updates the error handling
to suppress the error thrown from jdbc when the monitor has detected a
query that is terminated by pg_terminate_backend.