lockAdd: case of 2D plane in 3D

[WeiqunZhang]

Summary

In HiPACE++, atomicAdd is used on 2d x & y planes even though AMREX_SPACEDIM is 3. In that case, we would have all threads competing for a single lock in the previous implementation of lockAdd. This PR fixes this use case by having locks associated with the y-direction when the number of cells in the z-direction is 1.

Additional background

Hi-PACE/hipace#1059

Checklist

The proposed changes:

• fix a bug or incorrect behavior in AMReX
• add new capabilities to AMReX
• changes answers in the test suite to more than roundoff level
• are likely to significantly affect the results of downstream AMReX users
• include documentation in the code and/or rst files, if appropriate

[AlexanderSinn]

I tested HiPACE++ running on two 48-core CPUs with 4095*4095*100 cells and a tile size of 256^2. While the lockAdd version from this PR is faster, particularly at lower thread counts, it does not solve the bad scaling past 48 threads.