Have adaptable flush sizes histogram buckets (proportional to the flushes memtables for that cf / reason)

[udi-speedb]

The "Flushes" top level object has nested objects per cf, and, every cf has nested objects per flush reason.
Per cf/flush reason, there is a histogram of the number of flushes per size of flush. At the moment, these sizes are fixes (please see attached snapshot).
The proposal is to adapt the buckets of the histogram to the sizes of the memtables that were flushed in this cf and for that flush reason.
For example, if the max flush size for a cf + flush reason is 100mb then use that to create the buckets:
First bucket up to - 100 * 0.01 . second up up 100 * 0.05 , third 0.2 etc