Use the isoband package to calculate contour lines and bands

[paleolimbot]

This is a PR to address #3044. While the desired behaviour (calculating contours at draw time that can be labeled) can't be done immediately because Geoms don't have access to non-position scales, a good start is to use @clauswilke's isoband package to calculate contour lines (as opposed to grDevices::contourLines()) in stat_contour(). The isoband package also has a function to return contour polygons, which are implemented in this PR as stat_contour_filled() and geom_contour_filled().

Additional refactoring steps were:

• The complete parameter is not used in the current calculation and hasn't been used for about 4 years (estimated from the git blame view), so I removed it.
• I added function parameters and documentation for bins, binwidth and breaks. These appeared in examples but were not function arguments and were not documented.

Things I am not sure about:

• The calculation of breaks in this way is different than the calculation of bins and breaks everywhere else in ggplot2, and I wonder if they should be updated in this PR. In particular, letting breaks be a function I think would be helpful.
• I'm not sure about how I implemented geom_contour_filled(). There is no GeomContourFilled, it just calls layer() with stat = "contour_filled" and geom = "polygon".
• I'm not sure what the corner cases are for isoband::iso(band|contour), but I think I should probably add more tests to make sure inappropriate/weird inputs are handled properly (such as for Proposal: Warning when using density2d + coord_map #2702).

[paleolimbot]

A few more notes:

• isoband output isn't stable in SVG (see Refactor geom_contour()? #3044 (comment) ), so it can't be visually tested. I've tested all the parts non-visually and added an independent test for polygons, which were being tested using stat_density2d() previously.

[clauswilke]

I had a quick look and didn't see anything that concerns me. I'll be traveling extensively over the next two weeks, starting today. Therefore, you probably shouldn't wait for a more careful review from me, unless there's anything you feel uncertain about.

[paleolimbot]

I appreciate you taking a look! When ggplot2 is ready for scale access at draw time, I imagine I will have many more questions!

[hadley]

While you're in here, it would be useful to quickly profile this code, just to see if there are any major bottlenecks we could improve.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/