[wasikarn]

What

• run_hook_tests() collects every test file with a non-zero exit and prints failing: <files> after the layer's output.
• The final gauntlet: FAILED message repeats that line, so the last lines of a CI log name the culprit.
• New tests/scripts/test-gauntlet-failed-summary.sh.

Why

A red gauntlet log never named the failing file. One CI failure log ran to 3284 lines in 49 sections, and the runner's /tmp, where the full logs are kept, is gone after the job.

Proof

• New test extracts run_hook_tests() with hook_test_files() stubbed to one passing and one failing fixture, and runs the script's closing if-block on a fixture log. Red before the fix (2/4 failed), green after (4/4).
• Mutation check: restored the original script, the test went red again; re-applied the fix, green.
• Exit codes are unchanged. bash 3.2 only (space-joined string, no arrays).
• test-run-gauntlet-wiring.sh and test-gauntlet-log-preservation.sh pass.
• audit.sh: 0 CRIT. run-gauntlet.sh: all layers passed.

Closes #387