fix duplicate primary check to not return 1 row per duplicate

[cliu587]

Currently the duplicate primary key check returns 1 row per duplicated primary key. For large tables, this can make the script go OOM and throw a non-helpful 'Script returned with exit status 137' rerror.

This change makes the primary key check script only return a single row containing the count of total duplicate primary keys.

[sb2nov]

Needs to be part of larger migration