Fix the rollback snapshot, which had never worked

The snapshot path was single-quoted inside the ssh payload, so the remote shell
never expanded the command substitution: every deploy wrote into one directory
named after the un-expanded literal, and the second deploy died on 'File
exists'. It looked correct for exactly as long as there had only ever been one
deploy — which is the failure mode this repo keeps finding, and the reason
deploy.sh smoke-tests the public hostname instead of trusting an exit code.

Two real snapshots now exist on the host.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019mt6sHQHEnEYrJZvoMCJSB

deploy/deploy.sh

@@ -29,7 +29,13 @@ ssh "$HOST" "set -e
   [ \"\$free\" -ge 3 ] || { echo \"only \${free}G free on /\"; exit 1; }
   mkdir -p '$SNAPS'
   if [ -d '$ROOT' ] && [ -n \"\$(ls -A '$ROOT' 2>/dev/null)\" ]; then
-    cp -al '$ROOT' '$SNAPS/\$(date +%Y%m%d-%H%M%S)'
+    # Double quotes, not single. Inside single quotes the remote shell never
+    # expands the command substitution, so every deploy wrote into a single
+    # directory named after the un-expanded literal, and the SECOND deploy died
+    # on 'File exists'. It looked like it worked for exactly as long as there
+    # had only ever been one deploy. Note this comment is inside a
+    # double-quoted ssh payload: anything it names gets expanded too.
+    cp -al '$ROOT' \"$SNAPS/\$(date +%Y%m%d-%H%M%S)\"
   fi
   ls -1dt '$SNAPS'/*/ 2>/dev/null | tail -n +11 | xargs -r rm -rf"