The metrics reported in the "limit report" embedded in our HTML output don't exactly match between legacy and Parsoid, with some off by 10x. Since there are used for DOS prevention and other things, we should investigate this and try to make them match more closely. If the limits are irreconcilable due to Parsoid's differing methodologies, we should lint any large pages which are likely to exceed limits on Parsoid to warn editors that the page will no longer be editable once Parsoid replaces the legacy parser.