pg_catalog: fix up contents of `pg_catalog.pg_proc`

Currently, when running `opr_sanity.sql` subtest of `pg_regress` test suite, we get lots of rows on queries that should be returning none, indicating that some illegal values / incomplete information is present. For now, in order to avoid unnecessary churn of the pg_regress diff file when adding new builtins, we'll skip `opr_sanity.sql` altogether.

Jira issue: CRDB-38448

Epic CRDB-60809

Start with the pg_catalog.pg_proc contents and the opr_sanity.sql subtest in the pg_regress test suite. Identify the illegal values or incomplete information producing rows where queries should return none, then verify the catalog corrections against the regression output. Done means opr_sanity.sql can run without the current failures or unnecessary diff churn.

Written by the indexing model from the issue text.