New gridder (again!)
The new gridder based on the "exponential of semicircle" kernel passes our tests and is at least twice as fast as the current one (there is probably still some optimization potential).
When/how should we merge this? I'd be happy if people test this as long as it is still on the branch, but this is a bit complicated since you also need to checkout the "new_kernel" branch of
@parras, any suggestions?