pgsql: Fix new test for CATCACHE_FORCE_RELEASE builds

Fix new test for CATCACHE_FORCE_RELEASE builds

Two of the isolation tests introduce by commit bc32a12e0db2 had a
problem under CATCACHE_FORCE_RELEASE, as evidenced by buildfarm member
prion. An injection point is hit ahead of what the test spec expects,
so a session goes to sleep and there's no one there to wait it up. Fix
in the simplest possible way, which is to conditionally wake the process
up if it's waiting. An alternative output file is necessary to cover
both cases.

This suggests a couple of possible improvements to the injection points
infrastructure: a conditional wakeup (doing nothing if no one is
sleeping, as opposed to throwing an error), as well as a way to attach
to a point in "deactivated" mode, activated later.

Author: Álvaro Herrera <alvherre(at)kurilemu(dot)de>
Reviewed-by: Mihail Nikalayeu <mihailnikalayeu(at)gmail(dot)com>
Discussion: https://postgr.es/m/202511261817.fyixgtt3hqdr@alvherre.pgsql

Branch
------
master

Details
-------
https://git.postgresql.org/pg/commitdiff/9e8fa05d3412d6adea453a8ec24d9a794498f4cf

Modified Files
--------------
.../index-concurrently-upsert-predicate.out | 23 +++-
.../index-concurrently-upsert-predicate_1.out | 116 +++++++++++++++++++++
.../expected/index-concurrently-upsert.out | 23 +++-
.../expected/index-concurrently-upsert_1.out | 116 +++++++++++++++++++++
.../specs/index-concurrently-upsert-predicate.spec | 18 ++++
.../specs/index-concurrently-upsert.spec | 18 ++++
6 files changed, 312 insertions(+), 2 deletions(-)