Right-hand rule polygons not followed with most recent matplotlib versions

[AugustinMortier]

Hello from Norway !
It looks like, since matplotlib 3.8.?, the geojson files computed with geojsoncontour are not always following the right-hand rule. This was probably taken care of by matplotlib before, but recent release notes suggest some "revert contour" deprecation.

[bartromgens]

https://pypi.org/project/geojson-rewind/ seems to be able to rewind geojson polygons. Could be used to fix this, or provide tips on how to fix this.

[bertcoerver]

When using geojsoncontour.contour_to_geojson on contours created by plt.contour from matplotlib>=3.8.0, different isolines (with same values) were connected together for my usecase:

After downgrading to matplotlib==3.7.3, the output from geojsoncontour.contour_to_geojson is normal again:

[maxdow]

They are some breaking changes concerning contourf method in the 3.8 release of matplotlib
https://matplotlib.org/stable/api/prev_api_changes/api_changes_3.8.0.html#contourset-is-now-a-single-collection

[toddscottik]

When using geojsoncontour.contour_to_geojson on contours created by plt.contour from matplotlib>=3.8.0, different isolines (with same values) were connected together for my usecase:

After downgrading to matplotlib==3.7.3, the output from `geojsoncontour.contour_to_geojson` is normal again:

I made a fix for it, here is my PR #26
Matplotlib still allows you to have an access to each closed contour in contour.allsegs field, while contour.get_paths() contains a single Path for each level. I understand, there is a room for improvement, maybe it's better to get only one feature for every level instead of one feature for each closed contour.