fix(bundle_resolver): keep filesystem fallback alive when mdfind times out

[sebastianbreguel]

Summary

Under set -euo pipefail, hit=$(run_with_timeout 2 mdfind ... | head -1) kills the shell when run_with_timeout returns 124 (timeout). That's exactly the case the timeout was supposed to survive — the resolver should fall through to its direct filesystem scan, not die.

Appending || true to both hit=$(...) assignments in bundle_has_installed_app preserves the fast-path attempt but lets both mdfind failures and timeout-driven 124 exits fall through to the Info.plist walk below.

Why this showed up now

This has been surfacing as a CI failure on macos-latest for PR #770 (and blocking #772 / #773 rebases). The failing test is bundle_has_installed_app finds an app by CFBundleIdentifier (Spotlight miss) — test #9 in the Unit & Integration Tests job. It's not a flake: mdfind can legitimately time out on hosted runners and trip exactly this path.

Repro without the fix:

set -euo pipefail
run_with_timeout() { return 124; }
hit=$(run_with_timeout 2 mdfind "x" 2>/dev/null | head -1)   # shell exits here
echo "never reached"

Test plan

• Added bundle_has_installed_app falls back after run_with_timeout returns 124 (set -e + pipefail regression) to tests/bundle_resolver.bats. Mocks run_with_timeout to return 124 and asserts the filesystem fallback still finds the app.
• Verified the new test fails on upstream/main (reproducing the CI failure) and passes with the fix.
• Full bats tests/bundle_resolver.bats green (8/8) locally.