Fixed that bridge angle is lost on shell layers

[nstbayless]

Previously, if a bridge was on a layer which had solid infill due to a horizontal shell, the bridge angle would be reset to -1. This could occur e.g. if a model has two bridges on on neighbouring layers (even if the bridges aren't near each other).

This may be a fix for issues #3175 and #3139.

[AppVeyorBot]

✅ Build Slic3r 1.3.0-master-2290 completed (commit e2914859b0 by @nstbayless)

[lordofhyphens]

@nstbayless I'd appreciate a regression test case and an example as situations where the existing behavior is doing something wrong before merging.

[VanessaE]

This does fix #3139, for the most part. Slic3r doesn't necessarily merge neighboring bridge areas together though -- for example, the bridging that tops-off the embossed text ("CT3D.xyz") on the bottom of a Benchy is still divided into many smaller patches.

While bridges are more correct now (like with Triffid Hunter's bridging test, or some of my models), Slic3r still chooses some odd/non-ideal bridging angles (but I think that's a different issue).

[lordofhyphens]

Seeing as this didn't cause as test failure before and does appear to be a bug, please add a test to
src/test/libslic3r/test_printobject.cpp

[lordofhyphens]

@nstbayless do you have any interest in completing this PR?

[nstbayless]

@lordofhyphens Sure thing. I'll have free cycles to implement the required changes in about 2 weeks soon.

[VanessaE]

*bump*

@nstbayless I know time is short but let's get this done and merged.

It fixes the general direction issue, but bridges are still being split in mid-air (or not being combined).

[VanessaE]

I just tested this against 4f5b935. It still seems to work as it did before: it's not perfect, but it compiles fine, all unittests pass, and it fixes bridging angle well enough to to be worth merging, in my opinion. Whatever details are left to fix should just be sorted out later.