Extract page-level CrUX object to top-level schema

[rviscomi]

Right now the page-level CrUX data can be queried using JSON_QUERY(payload, '$._CrUX'). Similar to our work to split up the custom metrics blob, we could extract the CrUX object to make it cheaper and faster to query.