revert: undo PR #3/#4 pending CI-only e2e-main-help regression investigation

[takeokunn]

Summary

• Reverts merge commits 7e4a9a2 (PR refactor: split builtin printf/macros and relocate highlight module #3, printf/macros split + highlight relocation) and 7839d19 (PR docs: fix stale define-builtin location in macro consolidation audit #4, docs fix dependent on PR refactor: split builtin printf/macros and relocate highlight module #3) back to the tree at 3efe546 (right after PR refactor: follow the cl-prolog-kit / cl-dataflow-kit rename #2).
• Confirmed via CI history: e2e-main-help-exits-cleanly passes reliably before PR refactor: split builtin printf/macros and relocate highlight module #3 and fails deterministically in every CI run after it (both x86_64-linux and aarch64-darwin), while passing 100% in two separate full local suite runs on the same commit. This points to a CI-environment-specific (resource/timing sensitive) regression in the process/job-control refactoring in PR refactor: split builtin printf/macros and relocate highlight module #3 that could not be pinned down with confidence without further investigation.
• git diff 3efe546 HEAD is empty, confirming this exactly restores the pre-PR#3 tree.

Test plan

• Tree diff against 3efe546 (last known-good commit) is empty
• CI green on this PR (expected: same baseline failures as 3efe546 — build release binary patchelf permission issue and occasional aarch64-darwin PTY flake — but NOT e2e-main-help-exits-cleanly)