This PR fixes an issue in chain indexer. Currently chain indexer will
validate whether the stored data is canonical by comparing section head
and canonical hash. But the header of the checkpoint may not exist in
the database. We should skip validation for sections below the
checkpoint.