The common way to include buildings in a viewshed analysis is to translate the buildings into raster, then add it to the original DTM to get a new DTM with building and then run the analysis on the new DTM.
The problem with this is that most DTM is low resolution (10 meter is good dtm for most areas).
The vector data is in better accuracy.
You need to resample you 10 meter raster into 1 meter raster (100 times bigger) to get good results.
Is there any other option? Do you think putting an idea in esri site to include polygon layer in the viewshed is a good idea?
