Revisit GROUP BY translation for BigQuery

Currently SqlRender translates GROUP BY statements by replacing the column names with column numbers:

```r
translate("SELECT f(a) AS my_value, COUNT(*) AS my_count FROM t GROUP BY f(a);", targetDialect = "bigquery")
# "select f(a) as my_value, count(*) as my_count  from t  group by  1 ;"
```


 As far as I can tell, this is no longer needed, and it would simply things a lot if we can remove it.