Distinct count takes too long for big tables.

[acautin]

It seems most of the time is spent finding if the values are already present here:
https://github.com/K2InformaticsGmbH/dderl/blob/master/src/dderl_fsm.erl#L1559

Evaluate if the performance can be improved by using ets tables or maps for accumulating the intermediate results without having a big impact in the memory consumption/garbage generated.