ENH: Make aggregation step in Q_and_F more flexible

[timmens]

Currently, LCM only supports the aggregation function

$$ H(u, V) = u + \beta \cdot V. $$

There should be the possibility to use a custom aggregation function, with some typical ones pre-implemented and selectable via an enum.

[hmgaudecker]

When tackling this (or just before), we should rename beta to discount_factor everywhere.