We found that for writers s in STATE_LOCKED_WAITING, the notify-one
function needs to be called, and the cost of calling this function is
very high especially when there are many writers that need to be
awakened. So, we Parallelize this progress.

To wake up each writer to write its own memtable, the leader writer
first wakes up the (n^0.5-1) caller writers, and then those callers
and the leader will wake up n/x separately to write to the memtable.
This reduces the number for the leader's to SetState n-1 writers
to 2*(n^0.5) writers in turn.

vcpu=160, benchmark=db_bench
The score is normalized:
| case name         | optimized/base |
|-------------------|----------------|
| fillrandom        | 182%           |
| fillseq           | 184%           |
| fillsync          | 136%           |
| overwrite         | 179%           |
| randomreplacekeys | 180%           |
| randomtransaction | 161%           |
| updaterandom      | 163%           |
| xorupdaterandom   | 165%           |