test(bash32): unify the two heredoc-in-substitution scanners after #644 and #653 merge #665

🤖: AI agent (First Mate), AI-assisted; human review required.

Problem

Two open pull requests each add a scanner for the same bash 3.2 hazard: a here-document inside $( ) or <( ) whose body has odd quotes, an unpaired backtick or unbalanced parentheses. Bash 3.2 fails at parse time on such a body; bash 4+ accepts it.

• fix(windows): port acq provider-input path to Windows/MSYS #644 adds test/bats/heredoc-scan.py (about 40 lines) and calls it from test/bats/151-heredoc-in-substitution.bats.
• feat(test): add local bash 3.2 test runner #653 adds scripts/scan-heredoc-subst.py (about 100 lines). It also handles arithmetic, case pattern heads, function heads and <<< here-strings, and it backs scripts/test-acq-bash32 --syntax.

Both flag the real failure that reached CI (the commit where a Python comment containing an apostrophe, backticks and parentheses broke the facts reader under bash 3.2). Both pass on the fixed file and on current main. Neither is a real parser; bash-3.2 -n is the authority. Having two copies means the rules will drift.

Proposed change

Keep scripts/scan-heredoc-subst.py as the single scanner (it is the superset). Point test 151 at it, delete test/bats/heredoc-scan.py, and keep test 151's positive and negative cases so coverage does not drop. The decision record added with #653 already says to fold the two into one file.

Trigger

Blocked until both #644 and #653 are merged to main. Pick this up after the second of the two merges.