Fix: gh_run exits prematurely under set -e

[codeforester]

Summary

gh_run in lib/bash/gh/lib_gh.sh is broken in scripts that use set -e (or set -euo pipefail).

Root cause

gh_run() {
    local status

    gh_require_cli || return 1
    gh "$@"          # ← unchecked: with set -e, a failing gh exits the script here
    status=$?        # ← never reached on failure
    ((status == 0)) && return 0
    gh_report_command_failure "$status" "$@"
}

gh "$@" is a plain command in the function body. With set -e active, any non-zero exit from gh causes the calling script to exit immediately — before status=$? is captured. gh_report_command_failure never runs, so the caller gets a silent abrupt exit instead of a controlled error with diagnostics.

The BATS suite doesn't expose this because capture_command does set +e before calling.

Fix

gh_run() {
    local status=0

    gh_require_cli || return 1
    gh "$@" || status=$?
    ((status == 0)) && return 0
    gh_report_command_failure "$status" "$@"
}

The || status=$? pattern keeps gh on the left side of ||, which is exempt from set -e.

Test gap

Add a test that calls gh_run from a script with set -euo pipefail active and a fake gh that returns non-zero, verifying that gh_report_command_failure output appears rather than a silent exit.