Consider a build-time check that test/install/*.sql sets ON_ERROR_STOP

[jnasbyupgrade]

Background

In Postgres-Extensions/cat_tools's test/install/load.sql (PR #54), \set ON_ERROR_STOP on
turned out to be load-bearing in a way that's easy to silently break: test/install/*.sql files
run in their own, self-comparing pg_regress entry (see pgxntool's own docs on the test/install
feature), so a RAISE EXCEPTION/hard-error check in one of these files only actually fails the
build if ON_ERROR_STOP is set -- without it, psql prints the error and keeps going, and nothing
ever diffs to catch it. Removing that one \set line silently turns every hard-error check in the
file into a no-op, with no test failure to reveal it.

Proposal (raised for discussion, not a confirmed ask)

Consider whether pgxntool itself should have a build-time check that every test/install/*.sql
file sets ON_ERROR_STOP before anything else that could raise an error -- e.g. a make lint-style
scan, or a check-test-install-error-stop target wired into the existing test/install feature's
own machinery. This would be generic to any pgxntool-based extension using test/install, not
cat_tools-specific.

Open question

Whether this is worth building at all -- an explicit check adds real complexity (parsing/scanning
every install file, deciding what "before anything that could error" means precisely) for a
problem that a code-review comment/convention might catch just as well. Recording the idea here
rather than committing to it; feel free to close as wontfix if the maintainers don't think it's
worth the complexity.