test: add bats tests for benchmark shell scripts

[BryanFRD]

Summary

• Add 19 bats-core tests covering compare.sh and format-release.sh
• Tests cover: regression detection, threshold comparisons, edge cases (missing baseline, empty results, zero values), delta calculations, pivot table output, multi-tool/multi-method handling
• Add test job to CI pipeline, gating both release and benchmark jobs

Closes #48

[Copilot]

Pull request overview

Adds bats-core test coverage for the benchmark reporting scripts and wires the test suite into CI to gate release/benchmark workflows.

Changes:

• Add bats tests for scripts/compare.sh (regressions, thresholds, missing baseline/latest, absolute limits, size regression).
• Add bats tests for scripts/format-release.sh (pivot formatting, deltas, multi-method/tool output, empty input, usage errors).
• Add a CI test job and make release/benchmark depend on it.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
tests/format-release.bats	New bats tests validating markdown formatting and delta rendering for release output.
tests/compare.bats	New bats tests validating regression detection behavior and threshold handling.
.github/workflows/ci.yml	Introduces a test job and gates release/benchmark on lint + tests.

[Copilot]

make_json already wraps the provided benchmark entries in { ... } via "benchmarks": {$benchmarks}. Passing '{}' here produces invalid JSON ({{}}). Call make_json with no additional benchmark args (or adjust make_json to accept a pre-wrapped object) when you want an empty benchmarks map.

[Copilot]

This call passes '{}' into make_json, which results in "benchmarks": {{}} (invalid JSON). For an empty baseline, invoke make_json "$TMPDIR/baseline.json" with no benchmark args (or update make_json to avoid double-wrapping).

[Copilot]

This call passes '{}' into make_json, which results in "benchmarks": {{}} (invalid JSON). Use make_json "$TMPDIR/baseline.json" with no additional args (or make make_json accept a raw JSON object) to produce an empty benchmarks map.

[Copilot]

The bats tests invoke scripts/compare.sh and scripts/format-release.sh, both of which hard-require jq and exit non-zero if it’s missing. The test job should install jq explicitly (or add a dedicated setup step) to keep CI deterministic across runner image changes.