This eliminates the leftover concept of "memtable-in-flush" aka "Opt
". As seen in https://github.com/ysono/pancake/issues/56 in the first diagram, the "memtable-in-flush" concept is applicable only to a very simple design for concurrent execution. This was never implemented, and now that we have secondary indexes, it is undesirable to implement it. Hence, I'm simplifying the code and aligning this serial execution impl with the design at the 3rd row of the first diagram there.
This eliminates the leftover concept of "memtable-in-flush" aka "Opt
". As seen in https://github.com/ysono/pancake/issues/56 in the first diagram, the "memtable-in-flush" concept is applicable only to a very simple design for concurrent execution. This was never implemented, and now that we have secondary indexes, it is undesirable to implement it. Hence, I'm simplifying the code and aligning this serial execution impl with the design at the 3rd row of the first diagram there.