Populate should leave evidence of success state

[CBroz1]

Problem: It is possible to run common_behav.StateScript.populate for a given key and have no resulting insert. This leaves previously attempted keys in the key_source, which will be re-attempted if re-run. The current structure doesn't distinguish between 'attempted' and 'not attempted'.

Proposed: By modifying the secondary key to be nullable, we can retain a record of which keys have been attempted. Optionally, another secondary key could be added to be explicit (e.g., has_assoc_file: bool). I have seen this pattern on other tables. I propose migrating all populates to retain records of attempted keys with null entries

Considerations: In this case the cost of rerunning is small, a couple conditionals, but this habit opens the door to unnecessary processing time elsewhere. Stylistically, lab members tend to populate a specific key as needed, which will not proc this issue, but datajoint documentation often uses bare populates, so use habits may change as more folks try out spyglass