- All2AllIterative: Use PropagateDownwind WS_eff as initial guess, determine unstable flow cases and limit WS_eff of idling WT in unstable flow cases to max previuos WS_eff
- Apply rotor average model on blockage deficit
- Speed up Rathmann and SelfSemilarity by implementing calc_layout_terms etc.

Reduce memory consumption of All2AllIterative with NOJDeficit + add test of memory usage and memory leak

Allow different superposition models for wake and blockage + add blockage to flowmap with weightedSum (was missing before)

implement support for i-dependent WS and WD in site and wt-dependent WS and WD kwarg to WindFarmModel.__call__

Reduce memory consumption in PropagateDownwind by doing inplace flattening of dw_iil, hcw_iil and dh_iil. Note that this may have side-effects in the parent functions

before default rotor center was shared among wind farm models, i.e. wakedeficit model of last instance was applied to all instances