Alter uprating functionality to prevent the need to manually declare each uprated variable in data/datasets/cps/cps.py

[anth-volk]

At present, in order to successfully allow a value to be uprated, it must be defined as something (at minimum, a zeroed-out array) within data/datasets/cps/cps.py; otherwise, it's returned as a series of NaNs, breaking any derivative formulas in the process. This came up recently with uprating applied to social security within policyengine-us, whereby social_security_retirement had been defined, and thus evaluated correctly, but social_security_survivors was not, and thus NaNed out.

Could also be valuable to add a ValueError throw at the same time to ensure that we can adequately trace these in the future, if necessary.