A cautionary note on the use of the frozen‐core approximation for correlation energy calculations involving alkali metals