testsuite: accept the -M local-transfer breakage in the backport skip list

fake-super-acl-xattr is the last test the 3.5.0 suite still fails on this
branch, and unlike the other entries it is not an absent feature: any
--remote-option on a local transfer makes the receiver write to the current
directory instead of the destination, silently and with exit 0.  It predates
every security commit here, stock 3.2.7 does the same when given -s, and
v3.4.1 and 3.5.0 are unaffected.

We are shipping that, so the test is excluded -- but in its own section, with
the defect described and pointing at the analysis, because silencing a real
failure in the same list as "this feature does not exist" is how it stops being
remembered.  Anything that reads this list needs to see the difference.

testsuite/skiplist/backport.txt

@@ -33,3 +33,20 @@ rrsync-no-overwrite-backup-collision
 rrsync-no-overwrite-delay-updates
 rrsync-no-overwrite-logfile
 rrsync-no-overwrite-partial-dir
+
+# --- KNOWN BUG, DELIBERATELY ACCEPTED -------------------------------------
+# Everything above is a feature or helper this branch does not have.  This one
+# is different: it is a REAL DEFECT in this branch that we have chosen to ship.
+#
+# Any --remote-option (-M) on a LOCAL transfer makes the receiver write to the
+# process's current directory instead of the destination -- silently, with exit
+# 0 and a normal transfer summary.  fake-super-acl-xattr is simply the only
+# test in the 3.5.0 suite that uses -M on a local copy, so it is the only one
+# that catches it.  Stock 3.2.7 fails the same way when given -s; v3.4.1 and
+# 3.5.0 are unaffected; the breakage predates every security commit on this
+# branch.  Full analysis, including the strace and what was ruled out, is in
+# dev-notes/v327-backport-remote-option-wrong-dir.txt.
+#
+# Remove this line if that bug is ever fixed -- it is the only thing watching
+# for it.
+fake-super-acl-xattr