We now query selector and take textContent synchronously. This
avoids any issues with async processing: node being recycled,
detached, etc.
More methods will follow with the same atomic pattern.
Drive-by: fixed selector engine names being sometimes case-sensitive
and sometimes not.
