Currently G1 only actually revises young gen length if the number of remembered set entries exceeds the estimated young gen length.

This is problematic because if the initial guess for remembered set length is too high, G1 never re-evaluates young gen length, or re-evaluates only if young gen length is already too high to meet pause time goals (i.e. way too late).

So instead of gating the young gen revising, unconditionally do it.