Don't throw out short polylines

[hpinkos]

Made the epsilon values smaller for Polylines, Corridors and PolylineVolumes

This forum post reported a problem with the PolylineVolumes

[mramato]

@bagnell does the relative epsilon stuff you put in a while ago come into play here, or is this really just a case of the tolerance not being high enough?

Related question: Do we always render the polyline if there are only two points or do we end up deduping to 1 point and then throwing it out?

[hpinkos]

@mramato we'll throw it out if the polyline only has two points and the points are almost the same

[hpinkos]

Fixes #3293

[pjcozzi]

Update CHANGES.md.

@bagnell can you please review?

[hpinkos]

@bagnell this is ready now.

[bagnell]

does the relative epsilon stuff you put in a while ago come into play here

@mramato Yes, it was changed to check the absolute or relative error. Here it would check the relative error. We need to go through and check all of the epsilon values, but I think we are just fixing them as issues come up.

[mramato]

Should there be a constant with a single epsilon defined somewhere that all of the geometry uses? Should that be user settable?

[mramato]

And is 10 the right answer, or just a random guess?

[hpinkos]

@mramato 10 is a random guess. It's probably small enough, but at some point we should probably figure out what's the smallest distance that's still worth rendering. 6 was definitely too big because you can still see things at that size even without being zoomed in all of the way.

Should there be a constant with a single epsilon defined somewhere that all of the geometry uses? Should that be user settable?

I don't think it would need to be user settable, but having a constant somewhere that all of the geometries use is a good idea. I added a note to #2364