Allow different engines in a partitioned table.
What is mostly needed is allow one to combine InnoDB with S3 tables but other combinations should be possible . InnoDB couldb be a OVERFLOW partition that handles all inserts while the other partitions are only used for read.
From optimizer point of view, we should be AND the flags from all engines
that is part of a query. If S3 and InnoDB is in the same query, we would
loose the clustered key cost optimization, however I think we can survive
To support multiple engines, we probably also need to do MDEV-5271