Fix for rectangle and square grid dimensions

[rowanwins]

So in digging into #2079 I found that we using the haversine distance formula for calculating how big to make the grid cells.

Where this falls down is where you bbox if > 180 degrees because it then finds the shortest distance, and the wider your bbox of interest the worse the effects would be.

// A straight line 
turf.distance([-180, -90], [180, -90], {units: 'degrees'})
==> 0

Swapping to a flat earth earth calc we do something like

180 - -180
==> 360

There doesn't seem to be an easy way to hack our distance calc to produce something similar (at least in my initial play) so adjusting the approach seems reasonable and gives more expected results when generating large grids (eg see https://github.com/Turfjs/turf/blob/master/packages/turf-rectangle-grid/test/out/big-bbox-500x100-miles.geojson)

Resolves #2079
Resolves #1948
Resolves #1598

The same lines of code possibly need to be ported to point-grid and triangle-grid. See #1886

[rowanwins]

Alternatively I need to look into how calculate a harversine distance going the long way around (when the bbox is >180), that seems achievable, although a bit fiddlier

[mfedderly]

It looks like this broke the tests in @turf/square-grid, they probably just need to be re-updated. Github also isn't letting me do the rich diff view for these geojsons, but I'm assuming you looked at the output and think its better?

[rowanwins]

I've checked the results of the proposed PR against QGIS and we now generate almost the exact same results.

[mfedderly]

This looks great to me