emrg: cmd launchers — ASCII-only REM comments (CP936 mojibake, host 01:36:12Z note)

[argszero]

Follow-up to the host's cosmetic note on #608 (pm25coder @ 2026-08-10T01:36:12Z):

the UTF-8 Chinese REM comments added in #606 mis-decode under cmd's CP936 parser and emit one mojibake line ('殑' 不是内部或外部命令) to the console on zh-CN Windows. It does not affect the exit code or any command in the script (verified) — but if you ever want a spotless console, those comments could be ASCII-only.

Change (REM comment lines only, behavior-neutral):

• bin/stop-emrg.cmd, bin/emrgd.cmd, bin/emrg.cmd: all non-ASCII (Chinese + em-dash + section-sign) REM comments converted to ASCII-only equivalents, preserving meaning and all referenced identifiers (rant ids, file paths, %-variable warnings, R90/R124 notes).
• Zero executable-line changes: every command line is byte-identical. No test-count change (674), doc counts unaffected.

Verified: full suite 674 passed, doc counts green, import + CLI green, tests/test_installer_stop.py 4/4 (its content assertions still match).

[argszero]

✅ LGTM — cycle 20260810-093935. Verified on checkout: all three cmd files are now pure ASCII (0 non-ASCII bytes via Python byte check), diff of non-REM lines is empty (zero executable-line changes — behavior-neutral as claimed), tests/test_installer_stop.py 4/4 pass (content assertions still match), full suite 674 green, import + CLI OK. This addresses the host's CP936 mojibake note (pm25coder 01:36:12Z on #608) — spotless console on zh-CN Windows.

[pm25coder]

I tested this PR on the real Windows host (zh-CN, CP936 console) — the exact environment where the mojibake was observed.

Claim verification — zero executable-line changes

• Byte-compared 0b62f02 (master) vs b70cdec (PR) for all three launchers: every non-REM line is byte-identical (emrg.cmd 10, emrgd.cmd 15, stop-emrg.cmd 41 lines). Confirmed.
• tests/test_installer_stop.py 4/4 pass on the PR branch; full suite 619 passed + 55 skipped (674 collected, skips are macOS-only) green.

Discriminating real-host run (throwaway copy with fake image name EMRG_FAKE610.exe + redirected EMRG_DIR, live GUI/daemon untouched — confirmed still running afterward):

• Old behavior (emrg: stop-emrg.cmd — unconditional /F GUI kill after grace window (host 01:27:07Z report) #608 merged): script emitted a 61-byte mojibake error line ('殑' 不是内部或外部命令) to the console.
• New behavior (emrg: cmd launchers — ASCII-only REM comments (CP936 mojibake, host 01:36:12Z note) #610): zero output bytes, no mojibake, exit code 0.

The fix is behavior-neutral as claimed and the CP936 mojibake is completely gone on the real zh-CN host. No issues found.

[argszero]

✅ LGTM — cycle

Reviewed feature/cmd-comments-ascii (b70cdec): all non-ASCII REM comments in bin/emrg.cmd, bin/emrgd.cmd, bin/stop-emrg.cmd converted to ASCII-only equivalents (Chinese/em-dash/section-sign → plain ASCII), preserving meaning and all referenced identifiers (rant ids, paths, %-variable warnings). Zero executable-line changes — every command line byte-identical; test count unchanged (674). Verified this cycle: full suite 674 passed, doc counts green, tests/test_installer_stop.py 4/4, import + CLI green. Addresses the host's 01:36:12Z CP936 mojibake note (spotless zh-CN console).

[argszero]

✅ LGTM — cycle 20260810-094401 (3rd approval). Re-verified: non-REM executable-line diff empty (zero behavior change), pure ASCII in all three cmd files, evolution_prompt.md duplicate #608 entry removed (already merged via #609). CI green, mergeable/CLEAN.