triangulate on transformed points

[SimonDanisch]

This finally fixes the long standing issue of GeoMakie plotting broken polygons for non affine transforms, since it gets triangulated before transformation.
Sadly, Makie isn't ready yet for the best solution:

1. transforming in the poly recipe and handing through the transformed points doesn't work, since axis and limits will not work for e.g. log10, since it will apply the transform again.
2. leaving the transformation + triangulation entirely to the backend. This will need a larger refactor to make polygon a backend primitive.

To still finally have correct polygons, this PR applies the transformation before triangulation, but still forwards the untransformed points to the backend, which will lead to transforming two times, only if transform_func != identity.

With this PR:

Master:

[MakieBot]

Compile Times benchmark

Note, that these numbers may fluctuate on the CI servers, so take them with a grain of salt. All benchmark results are based on the mean time and negative percent mean faster than the base branch. Note, that GLMakie + WGLMakie run on an emulated GPU, so the runtime benchmark is much slower. Results are from running:

using_time = @ctime using Backend
# Compile time
create_time = @ctime fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @ctime Makie.colorbuffer(display(fig))
# Runtime
create_time = @benchmark fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @benchmark Makie.colorbuffer(fig)

	using	create	display	create	display
GLMakie	3.45s (3.40, 3.48) 0.03+-	529.48ms (511.20, 549.62) 15.26+-	484.92ms (475.85, 491.75) 5.74+-	7.77ms (7.52, 8.13) 0.22+-	26.03ms (25.84, 26.43) 0.20+-
master	3.46s (3.42, 3.52) 0.03+-	531.81ms (517.88, 550.02) 10.92+-	496.51ms (485.57, 527.01) 14.09+-	7.82ms (7.62, 8.19) 0.20+-	26.06ms (25.80, 26.33) 0.19+-
evaluation	1.00x invariant, -0.01s (-0.32d, 0.56p, 0.03std)	1.00x invariant, -2.33ms (-0.18d, 0.75p, 13.09std)	1.02x invariant, -11.59ms (-1.08d, 0.08p, 9.91std)	1.01x invariant, -0.06ms (-0.27d, 0.62p, 0.21std)	1.00x invariant, -0.03ms (-0.18d, 0.75p, 0.19std)
CairoMakie	2.87s (2.83, 2.95) 0.05+-	315.77ms (310.34, 327.99) 5.79+-	142.09ms (139.11, 147.97) 3.03+-	7.34ms (7.19, 7.49) 0.10+-	602.75μs (594.06, 610.70) 6.76+-
master	2.79s (2.75, 2.83) 0.03+-	316.90ms (313.31, 320.24) 2.95+-	141.51ms (139.85, 142.66) 1.10+-	7.41ms (7.30, 7.52) 0.07+-	597.74μs (593.98, 603.25) 3.03+-
evaluation	0.97x slower X, 0.09s (2.30d, 0.00p, 0.04std)	1.00x invariant, -1.13ms (-0.25d, 0.66p, 4.37std)	1.00x invariant, 0.58ms (0.25d, 0.65p, 2.06std)	1.01x invariant, -0.07ms (-0.74d, 0.20p, 0.09std)	0.99x invariant, 5.01μs (0.96d, 0.11p, 4.90std)
WGLMakie	3.63s (3.53, 3.72) 0.07+-	543.67ms (521.97, 560.69) 15.77+-	8.86s (8.63, 9.02) 0.15+-	9.75ms (9.42, 10.50) 0.36+-	72.07ms (67.02, 80.36) 4.31+-
master	3.56s (3.52, 3.65) 0.05+-	532.12ms (522.96, 546.91) 9.10+-	8.71s (8.60, 8.99) 0.15+-	9.60ms (9.43, 9.90) 0.16+-	71.07ms (67.66, 81.48) 4.86+-
evaluation	0.98x invariant, 0.07s (1.05d, 0.08p, 0.06std)	0.98x invariant, 11.55ms (0.90d, 0.13p, 12.43std)	0.98x invariant, 0.15s (1.01d, 0.08p, 0.15std)	0.99x invariant, 0.15ms (0.52d, 0.36p, 0.26std)	0.99x invariant, 1.0ms (0.22d, 0.69p, 4.59std)