Fix trapezoid construction for negative phi and older Geant4

[sethrj]

Thanks to @mrguilima for pointing out that there was an error in the calculation.

Also fixes https://github.com/celeritas-project/celeritas/pull/1223#discussion_r1592931865 .

[sethrj]

@mrguilima I've removed the validation checking from phi:

• Any value is valid and just wraps around as one would expect from sin/cos.
• Unlike InfWedge, the value is used to construct the shape as opposed to actually being stored on the shape.

If the value were part of the shape's data, we should regularize it so that there's a one-to-one correspondence between values and shapes produced. However, since phi just gets preprocessed into the same points, there's no need to impose extra effort on the user (e.g. it's ok if they want to specify the rotation as 3/4 turn instead of -1/4 turn).

[sethrj]

Is the proposed removal of validation satisfactory @mrguilima ?

[mrguilima]

Yes, it looks good! Thanks