Exit -c and -f with psql's status codes

[DiegoDAF]

Description

pgcli -c "select nope" exits 0, so a shell script or a CI job that calls pgcli cannot tell a failed statement from a good one. This makes -c and -f exit the way psql does.

psql's rule, measured with psql 18 against a live server: the exit status is that of the last -c or -f that ran. A failed -c exits 1. A script exits 0 even with errors in it, unless ON_ERROR_STOP is set; then the first failure stops everything after it, and a failed script exits 3. pgcli's on_error = STOP (the default) plays the part of ON_ERROR_STOP, and RESUME plays psql's default:

	psql	pgcli before	pgcli after
-c "select nope"	1	0	1
-c "select 1" -c "select nope", RESUME	1	0, second -c runs	1
-c "select 1" -c "select nope" -c "select 2", RESUME	0, all run	0, select 2 skipped	0, all run
-f err.sql, STOP / ON_ERROR_STOP	3	0	3
-f err.sql -f ok.sql, RESUME	0, both run	0, ok.sql skipped	0, both run
-c "select nope" -f err.sql, STOP / ON_ERROR_STOP	1, file skipped	0, file runs	1, file skipped

The last rows are the same loop's other departures from psql: under RESUME a failure kept the next -c or -f from running, and under STOP a failed -c did not stop the files.

One difference remains and is left alone here: pgcli runs every -c before every -f, while psql follows the order on the command line.

Tests: nine parametrized cases mirroring the measured psql runs, plus two through a real connection (a failed -c exits 1, a failed script exits 3). Ten of the eleven fail on main; the remaining case already matched psql. The two behave scenarios that run an invalid query asserted pgcli exits successfully, which wrote the bug down as expected behaviour (that assertion came with the original -c/-f PRs, mine); they now assert an error exit, and fail without this change.

Checklist

• I have added this contribution to the changelog.rst.
• I have added my name to the AUTHORS file (or it is already there).
• I installed pre-commit hooks (pip install pre-commit && pre-commit install).
• Please squash merge this pull request (uncheck if you would like us to merge as multiple commits)

On the third box: .pre-commit-config.yaml pins an older ruff-pre-commit than the one tox -e style installs, and the two disagree on formatting, so I ran ruff check and ruff format matching the CI instead, both clean.

[DiegoDAF]

The red builds are test_format_array_output_expanded only, the same in every job: cli_helpers 2.15.2, released today, aligns expanded output by display width, and that test has the old alignment written down. It fails on main too and is fixed in #1653; the other 2804 tests, including the new ones here, pass.