Fix download of documents at old schema versions (#2541)

Summary:
filterDocumentInPlace() runs raw SQL against the copy of a document
being downloaded, and two of its steps use columns that an old enough
document does not have yet. Either one fails the download with a
SQLITE_ERROR.

Now we check for the columns. An old document is left as it is, and
gains what it lacks when it is next opened and migrated.

Some filter options were getting repetitive across the app and tests, so
a small refactor is included.