Contest models highlight inherent inefficiencies of scientific funding competitions