Fix dangling nodes in ringsector

[Da-Be-Ru]

If the ringsector in commongrids.jl is called as a fullcircle, the final iarc iteration pushes nrad points into the coord array, but rightly skips setting the CellNodes for those so we get orphaned dangling nodes not assigned to any cells.

This causes some issues downstream which would be ironed out with these modifications suggested by @chmerdon

Additionally, setting the coordinate value type Tv to the standard Float64 stabilizes the cell lookups for gFindLocal! properly (fixing the excessively large runtimes of lazy_interpolate!).

[j-fu]

Thanks for the catch!
We should have tested for this long before.
Any ideas for more consistency checks ? If yes, make a new PR. Lets not do this here in order to get this finished quickly.