CEA should support more accurate partial adjacency in radiation geometry generation #3899
Description
Motivation:
Partial adjacency can help CEA distinguish different facade types more accurately (through radiation calculation), which will improve thermal simulation results and LCA results.
Problems in implementation
Currently in CEA's radiation calculation, there's only one intersection flag generated by the whole surface. If the midpoint of the surface is also close to another building geometry, the intersection flag will be set to 1, and no window will be generated on the surface (if it's a wall).
Then, the surface is subdivided, and sensors are generated from the subdivided surface center. Each sensor also has an intersection flag, but currently they are just copies of the original surface's intersection flag. This means that if a building's surface midpoint is covered by another building, the whole surface will be considered as "internal", regardless how the rest of surfaces would be. For example, in the following picture, all four highlighted surfaces will be internal walls or floors.
I suggest that the adjacency should be calculated on the sensor level or at least let user decide, who would like to spend more time in return of more accurate results. The user can also control how accurate they want their adjacency calculation to be by setting the grid size in radiation calculation.
Problem regarding candidate adjacent buildings
In some cases, for example when a building's dimension is simply too large or in a weird shape that results in a too large bounding box, adjacent buildings won't be accounted as adjacency. CEA accounts for potential adjacent buildings by measuring if the southwest corner of two bounding boxes are closer than 100m (implemented in cea.resources.radiation.geometry_generator.calc_building_geometry_zone and in cea.resources.radiation.geometry_generator.are_buildings_close_to_eachother).
In the following example, the apparently adjacent building would not even be accounted for and result in inaccuracy.
