bashunit bench exits 0 when a benchmark file fails to source or its set_up_before_script fails

[Chemaclass]

🤔 Background

bashunit bench prints a file-level failure and then exits 0, so the error reaches a human reading the log and never reaches CI.

Same scenario, both subcommands — one good file plus one whose set_up_before_script fails:

$ bashunit bench benchmarks/
✗ Error: Set up before script
bench_works 1 1 6
EXIT=0 <- bench

$ bashunit tests/
EXIT=1 <- test

A benchmark file with a syntax error is worse, because nothing is reported as failed at all:

$ bashunit bench benchmarks/
benchmarks/b_bench.sh: line 3: syntax error near unexpected token `{'
Running benchmarks/b_bench.sh
Bench ok [1/1] 6 ms
EXIT=0

Every bench_ function after the syntax error is silently absent from the table, and the run is green. test exits 1 for the same file shape.

💡 Cause

Two gaps:

1. bashunit::runner::load_bench_files (src/runner/bench.sh) sources with no status check. The test loop in src/runner/discovery.sh checks the status and matches the captured stderr for syntax error / unexpected EOF, because bash reports a syntax error and carries on with a zero status.
2. cmd_bench (src/main/run.sh) exits non-zero only for "No benchmarks found" and a baseline regression. It never consults the failure counter, so the existing set_up_before_script failure path — which already calls add_tests_failed — could not affect the verdict. It only looked correct when the failure left nothing to run, so "No benchmarks found" fired instead.

A benchmark function that merely returns non-zero should keep exiting 0: a benchmark measures time and has no assertion concept.

💡 Found by

Applying the pathological-fixture sweep from #1295/#1297/#1299/#1301 to the bench subcommand, which has its own runner and its own reports.

[Chemaclass]

Fixed in #1304 (merged).