perf(coverage): single-pass file scan + native regex in extract_functions

[Chemaclass]

Background

Follow-up to #636 / #643. The previous PR collapsed the seven grep subshells inside bashunit::coverage::is_executable_line into one. Three optimisation suggestions from @objctp's review remained open:

1. report_lcov re-scanned the coverage data file once per source line via get_line_hits (O(lines × data)).
2. report_text, report_lcov, report_html independently recomputed the same per-file executable + hit counts, each doing two extra file scans.
3. extract_functions spawned echo | sed / echo | grep subshells per line.

Changes

• New bashunit::coverage::compute_file_coverage <file> performs a single source-file pass, resolving hit counts via the already-single-pass get_all_line_hits helper. Returns executable:hit.
• get_file_stats, report_text, report_html now route through compute_file_coverage so per-file executable + hit counts are computed once instead of via two independent file scans plus repeated coverage-data greps.
• report_lcov reads get_all_line_hits once into a sparse indexed array and emits DA: entries while tallying executable + hit counts in the same single pass — no more per-line grep over the coverage data file.
• extract_functions replaces echo | sed -nE with bash native [[ =~ ]] + BASH_REMATCH, and replaces the per-line echo | grep -c '\\{' / '\\}' subshells with reuse of the already-computed brace-count parameter expansions.