Recoding an entire data frame using label-value pairs stored in a different dataframe

Suppose I have a dataframe:
df