Add midpoint to circular and elliptical arcs

[w0rm]

Closes #51

• midpoint for Arc2d and Arc3d take their respective types, because parametrisation for those doesn't exist.
• The test for EllipticalArc2d was flaky due to precision, I reduced it to 1.0e-9 and ran quite a few times, seems to always pass for me.

[ianmackenzie]

Looks good! I think the EllipticalArc2d test was getting occasional failures because the arc length parameterization is only guaranteed to be accurate to within the maxError specified when constructing it (in this case, meters 1.0e-3).

I think it would be most correct for the point2dWithin tolerance to be equal to the arc length parameterization tolerance; maybe both 1e-3 or perhaps both 1e-6 (some reasonably small tolerance that still lets the tests run quickly - the arc length parameterization is more expensive to construct if the maxError is smaller).

[w0rm]

Changed the precision to be 1e-3. 1e-6 didn't work because of FATAL ERROR: CALL_AND_RETRY_LAST Allocation failed - JavaScript heap out of memory

[ianmackenzie]

Huh, that's a little bit concerning that we run out of memory when trying to get an arc length parameterization accurate to 1e-6 meters on what's probably a several-meter-long elliptical arc...I mean that's a pretty high precision, but I would have thought it would be possible. Something to look into, but not worth worrying about as part of this PR.

One very minor thing I forgot to mention earlier: I think Quantity 1.0e-3 in the test could just be meters 1.0e-3, which would be more consistent, more type safe and would avoid needing the import line change.

[ianmackenzie]

Looks good! Just one more thing I just remembered - add yourself to the AUTHORS file =)