spherical subset false overlaps

Overlaps are detected by overlap_check_geant4 but present in neither overlap_check_tgeo nor visually in jsroot. The branch sphere-issue contains a minimal example of a spherical subset and a box, illustrating the issue. The relevant commit is 791792bc. Apologies, I do not know how to trigger the CI for this commit; the geometry root file is attached instead: detector_geometry.root.

The sphere subset is delimited by theta and phi limits, and its placement is rotated around the y axis by yrot. Depending on the settings of thetamax and yrot, false overlaps may appear. This commit has settings such that there are false overlaps.

Here is a table of various thetamax and yrot settings, and whether they pass or fail the overlap checks. The units of thetamax and yrot is degrees, and overlap_visual_jsroot indicates whether the overlaps are visually obvious when viewed in jsroot. Two rows commented with "issue" are the problems: the issue seems to only appear when we make a spherical subset and then rotate it during placement. It is possible this issue is in dd4hep itself.

thetamax	yrot	overlap_check_tgeo	overlap_check_geant4	overlap_visual_jsroot	comment
45	0	pass	pass	pass	expected behavior
45	90	fail	fail	fail	expected behavior
45	110	pass	fail	pass	issue
120	0	fail	fail	fail	expected behavior
120	110	pass	fail	pass	issue; this version committed