📝 Provide a description of the improvement

Model-to-view position mapping is heavily used, especially during editor initialization, when the whole model is downcasted after being loaded, and also in editor.getData().

Currently, Mapper recalculates the mapping for each single request. As a result, we do calculations for the same inputs multiple times, additionally, we don't use earlier, partially calculated results (e.g. if we calculated position for model offset 8, if we ask for model offset 9, we calculate again from the beginning).

Introducing some caching mechanisms will greatly improve performance.