fix(setup): plist-merge code-quality follow-up

Three issues raised by the code-quality reviewer on 0fd1838:

1. mergeSystemdEnv now early-returns when the template has no
   Environment= anchor, instead of silently dropping preserved lines
   (defensive guard for a future template change).
2. Both parsers (parsePlistEnv, parseSystemdEnv) now Buffer-normalise
   their input, removing the TypeError-vs-coerce asymmetry.
3. Two idempotency tests added (calling merge twice on the same
   input/output pair must be stable).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

test-features.mjs

@@ -564,6 +564,16 @@ test("mergeSystemdEnv first-install returns template unchanged", () => {
   assert.equal(mergeSystemdEnv("", SAMPLE_TEMPLATE_SYSTEMD), SAMPLE_TEMPLATE_SYSTEMD);
 });
 
+test("mergePlistEnv is idempotent", () => {
+  const r1 = mergePlistEnv(SAMPLE_EXISTING_PLIST, SAMPLE_TEMPLATE_PLIST);
+  assert.equal(mergePlistEnv(r1, SAMPLE_TEMPLATE_PLIST), r1);
+});
+
+test("mergeSystemdEnv is idempotent", () => {
+  const r1 = mergeSystemdEnv(SAMPLE_EXISTING_SYSTEMD, SAMPLE_TEMPLATE_SYSTEMD);
+  assert.equal(mergeSystemdEnv(r1, SAMPLE_TEMPLATE_SYSTEMD), r1);
+});
+
 // ── Cleanup ──
 closeDb();