Add syncthreads to scans prevent overwriting shared memory of previous chunk

[ivogabe]

Description

This PR fixes an issue that we had for a long time. Programs with scans would sometimes give incorrect values. As I found out, this is caused by missing synchronization in thread groups in the scan kernels.

Motivation and context

A thread group in a scan kernel may handle multiple chunks of the input.
A __syncthreads was missing between the end of one chunk and the start of the next, which caused that
shared memory could be overwritten in in the next iteration, before all threads are finished reading data from the previous iteration.
scanBlockShfl does perform synchronisation between its writes and reads, but there was no synchronisation between the reads of some iteration of the outer loop, and the writes in the next iteration.

How has this been tested?

Our quickhull implementation now works correctly on the GPU, whereas it previously would produce garbage values or crash (as it continues to compute with those garbage values). This was tested on an A100 and 4090.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have added tests to cover my changes.
• All new and existing tests passed.