increase swap_size_mb to 16GB in gha_lvm_overlay.sh to prevent memory issues during codeserver build (#2356)

turns out 16GB is probably overkill; the 256MB was not sufficient,
given the OOM kill, but I was later adding vm mem even when the problem
was insufficient nodejs heap limit, and not vm memory

```
./ci/build/build-vscode.sh: line 38:  9100 Killed                  npm run gulp "vscode-reh-web-linux-x64${MINIFY:+-min}"
Error: building at STEP "RUN ./get_code_server_rpm.sh && touch /tmp/control": while running runtime: exit status 137
ERROR:sandbox.py:Failed to execute process, see errors logged above ^^^
```

Author: jiridanek

ci/cached-builds/gha_lvm_overlay.sh

@@ -10,7 +10,8 @@ set -Eeuo pipefail
 # root_reserve_mb=2048 was running out of disk space building cuda images
 root_reserve_mb=4096
 temp_reserve_mb=100
-swap_size_mb=256
+# compilation from sources needs memory, for now that's codeserver
+swap_size_mb=16384
 
 build_mount_path="${HOME}/.local/share/containers"
 build_mount_path_ownership="runner:runner"