perf: reduce test-suite execution time (tracking)

[Chemaclass]

Tracking: reduce test-suite execution time

Running the full suite currently takes >4 min sequentially and ~48s with --parallel on a modern Mac. Profiling (macOS, default bash 3.2.57, Apple Silicon, commit 4d80e7c) shows most of that is framework overhead, not test logic:

Measured baseline

• 100 trivial tests (assert_true true) take ~5.0s, i.e. ~48ms of pure framework overhead per test. Across ~1160 tests that is ~56s before any real test work.
• Per test the framework pays ~9 external-binary forks (~21ms) plus ~12-14 $(...) subshells (~5ms):
  • 2x perl forks for clock reads (src/clock.sh:88, called from src/runner.sh:1070,1097) because bash 3.2 has no EPOCHREALTIME and macOS date has no %N. ~2400 perl processes per suite run.
  • 3x base64 pipelines in bashunit::state::export_subshell_context (src/state.sh:239-246), even when title/hook message are empty.
  • 1x grep -B 2 | sed data-provider probe per test (src/helpers.sh:310-314), re-reading the test file from disk each time.
  • 1x base64 -d decode (src/runner.sh:1110) and 1x rm -rf glob cleanup (src/globals.sh:70) per test.
• Acceptance tests fork ./bashunit as a subprocess 258 times (52 of 59 acceptance files). Each is a full cold start whose inner tests pay the same per-test tax again. This dominates the remaining sequential minutes.
• --parallel spawns with no default job cap (BASHUNIT_PARALLEL_JOBS=0; bashunit::runner::wait_for_job_slot short-circuits at src/runner.sh:293), so ~48s is essentially total CPU work / cores plus oversubscription overhead.

Child issues

Suggested order (independent, but small and mechanical first):

• perf(state): avoid base64 forks for empty title and hook message in export_subshell_context #762 perf(state): avoid base64 forks for empty title and hook message in export_subshell_context
• perf(runner): cache data-provider lookup per file instead of grep+sed per test #763 perf(runner): cache data-provider lookup per file instead of grep+sed per test
• perf(runner): remove avoidable subshell forks from the run_test hot path #764 perf(runner): remove avoidable subshell forks from the run_test hot path
• perf(clock): skip per-test clock reads when the clock impl forks an interpreter #765 perf(clock): skip per-test clock reads when the clock impl forks an interpreter (needs ADR, user-visible)
• perf(parallel): detect CPU count and benchmark a default --jobs cap #766 perf(parallel): detect CPU count and benchmark a default --jobs cap (measure-first)

Expected combined effect on this machine: sequential >4min toward ~2-2.5min, parallel ~48s toward ~25-30s. Re-run the shared benchmark after each merge and update this issue.

Constraints for all work

• Bash 3.0+ compatibility is non-negotiable (no declare -A, no [[ ]], no ${var,,}, no negative array indexing, no &>>). The perf target machine IS bash 3.2 on macOS.
• TDD, quality gate (./bashunit tests/, ./bashunit --parallel tests/, make sa, make lint, shfmt -w .) on every PR.
• One issue per PR; no batching unrelated changes.

Shared benchmark protocol

Run before and after each change, report numbers in the PR:

mkdir -p /tmp/bashunit-bench
for i in $(seq 1 100); do echo "function test_noop_$i() { assert_true true; }"; done > /tmp/bashunit-bench/noop100_test.sh
time ./bashunit --no-parallel /tmp/bashunit-bench/noop100_test.sh   # baseline: ~5.0s

# fork census (external binaries per run)
PS4='+X ${BASH_SOURCE##*/}:$LINENO ' bash -x ./bashunit --no-parallel /tmp/bashunit-bench/noop100_test.sh \
  2>/tmp/bashunit-bench/trace.txt >/dev/null
grep -oE ' (perl|base64|grep|sed|mktemp|rm|date) ' /tmp/bashunit-bench/trace.txt | sort | uniq -c

Also considered and NOT filed for now: lazy-sourcing large optional modules (coverage.sh, learn.sh) at startup. Sourcing all of src/ measures only ~28ms, so the win across 258 nested cold starts is ~5-8s and it complicates the single-file release build. Revisit if the cheaper wins land and cold start is still the bottleneck.

Line references are valid at commit 4d80e7c and will drift.