Constraining variance components to be positive in lme4 to avoid singular fit?

Question

I am using lmer in lme4 to model a response in several designed experiments with treatments randomly assigned to plots within blocks, and repeated measurements taken over multip