IVS-158 planar ifcface

[aothms]

Used mpmath (multi precision math, to operate on higher accuracy that 64bit floats) for additional precision because the typical tolerance like 1.e-8 gets rather close to the expected accumulated rounding errors for the point projection and normal estimation. (That's the idea after all, that's why we need the tolerance).

Only the outer bound is used for estimating the plane. The inner bounds need to agree to that, because of:

One loop is optionally distinguished as the outer loop of the face. If so, it establishes a preferred way of embedding the face domain in the plane, in which the other bounding loops of the face are inside the outer bound

https://ifc43-docs.standards.buildingsmart.org/IFC/RELEASE/IFC4x3/HTML/lexical/IfcFace.htm

We should probably also add a rule for:

NOTE If the IfcFace is used within an IfcFacetedBrep, where all faces are implicitly planar and having a disctinct outer bound, exactly one of the bounds shall be of the type IfcFaceOuterBound.