Fix CairoMakie heatmap filtering

[jkrumbiegel]

Description

Heatmap filtering was broken again because it used Cairo.FILTER_BEST, which does nothing in bitmap backends. This PR fixes that and reorders the logic a bit, so it should be easier to follow why the surface or rectangle path is chosen.

Type of change

• Bug fix (non-breaking change which fixes an issue)

Checklist

• Added unit tests
• Added reference image tests for new plotting functions, recipes, visual options, etc.

[SimonDanisch]

Why not leave this in a function?

[jkrumbiegel]

Because the flag logic didn't work anyway, there's no point in switching between the three, there's only two basically. So the main logic isn't about picking the flag but deciding whether to take the fast path or not.

[MakieBot]

Compile Times benchmark

Note, that these numbers may fluctuate on the CI servers, so take them with a grain of salt.

using time

master:  9.65 < 10.33 > 10.76, 0.35+-
pr:      10.15 < 10.39 > 10.95, 0.29+-
speedup: 0.93 < 1.00 > 1.02, 0.04+-
median:  +0.62% => invariant

This PR does not change the using time.

ttfp time

master   29.12 < 29.97 > 32.78, 1.23+-
pr       28.05 < 29.47 > 30.35, 0.80+-
speedup: 0.98 < 1.02 > 1.08, 0.04+-
median:  -1.67% => invariant

This PR does not change the ttfp time.