Update Pairplot example and spacing

[canyon289]

Fixes #212

Example Plots, It's easy enough to change the spacing if someone has a strong preference.

One thing to note is that there's some strange behavior with the y axis label positions. If this is problematic I create an issue.

[ahartikainen]

Can we split varname to multiple lines?

Dim first, then coords?

Something like

"\n".join(dim_name, coord1, coord2)

Also, will pairplot fail if gridspec (gs) is defined? Is there indentation error?

[ahartikainen]

Also, should divergent samples have greater contrast to main color?

E.g. a bit lighter color like salmon, lime, yellow?

[aloctavodia]

Splitting varnames to multiple lines would be great.
Default colors should be restricted to the default cycle (C0, C1, etc), as they have been optimized to be colorblind-friendly (C4 is yellow).

[ColCarroll]

Looks nice! We could also add a hexgrid version of this to the examples.

Splitting varnames is a good idea - I think make_label is used uniformly throughout, so it might even easy.

[ahartikainen]

Hex grid might be ok, but 2D KDE even better (with contours). See examples here https://agilescientific.com/blog/2017/12/14/no-more-rainbows

To get a working 2D KDE with fast_kde needs careful settings with limits.

[aloctavodia]

One good thing of hexgrid is that is really fast, we should also have 2D KDE (thus kind will accept 3 options 'scatter', 'hexgrid' and 'kde'). I may have a 2D-fast_kde version somewhere (not sure), I should check... Probably we should have a fast_kde and a separated fast_kde_2d version, in order to not slow down fast_kde.

[ahartikainen]

Yes, separate functions is needed. I have one working function somewhere.

[ahartikainen]

Nope, didn't find it, but basically it is a fork from 1D with specific limit constrains. Faststats has one implementation, but that could fail, if limits are outside min,max of data

[aloctavodia]

OK, don't worry I will send a PR on Monday or Tuesday.