Use adaptive_grid on endpoints

[BeastyBlacksmith]

I have ported the test-functions mentioned in JuliaPlots/Plots.jl#621 (comment) and as far as I can see there is no visual difference between including the endpoints or not. But my pitchfork example shows that plotting different functions which start or end at the same point are definitely better off by including the endpoints.

This addresses #63, though the concerns of respecting log-scale and y-limits are still valid.

Failures on nightly seem unrelated to me.

[mkborregaard]

I think the main issue is that you'll often want to pass 0 as an endpoint to see asymptotic behaviour of a function not defined at 0 - such as log or sqrt. Any suggestion for how to deal with that?

[BeastyBlacksmith]

that is already handled by the algorithm somehow, at least it doesn't error for me. It seems when plotting the NaNs are just skipped and the first interior point is used then. However, adjusting the ylimits might not give what you want. E.g.: plot(log,0,10, ylimits = (-100,2)) still has the first interior point at y ~= -6 for me. But that is a different issue.
And sqrt is well defined at 0.

[BeastyBlacksmith]

I could fix #48 in this PR as well, if wanted.

[mkborregaard]

That'd be great, just remember that it will be breaking on Plots and other packages that use adapted_grid (which I think is only Plots), so we'd need to up dependencies and push new minor versions.
Let me know when it's otherwise ready for review.

[BeastyBlacksmith]

should we also add a Project.toml?

So I will prepare a PR for Plots as well.

[mkborregaard]

Yes please :-)

[mkborregaard]

@goretkin does this fully address your concerns in #63 ?

[goretkin]

I don't think it does quite yet. I think you want to be able to refine the first and last intervals as well:

PlotUtils.jl/src/adapted_grid.jl

Lines 54 to 57 in 381fe0e

	# Skip first and last interval
	for interval in 2:n_intervals-1
	p = 2 * interval
	if n_tot_refinements[interval] >= max_recursions

[BeastyBlacksmith]

@mkborregaard This is ready for review. You need also my Plots-branch from JuliaPlots/Plots.jl#2204 to use the Plots interface.
@goretkin can you check, if this works for your function? Maybe add that to the list of functions in the test-directory.

[mkborregaard]

I don't have anything to add - impressed by the test suite, which seems to cover all the right corners.