[BUG] Memory issues when preparing large rasters for spatial multiprocessing

[katieb1]

Describe the bug
When working with large rasters, the 1 - Prepare Multiprocessing pipeline stage sometimes fails with an Unable to allocate memory error.

This is being caused by inefficiencies in line 195:
smp_grid = smp_grid.astype(int) * mask

To Reproduce
Steps to reproduce the behavior:

1. Download Linnea's test library
2. Try to run first pipeline stage

Expected behavior
The stage to run without error.

Software versions:

• SyncroSim: [e.g. 3.0.9]
• WISDM [e.g. 2.1.1]

[katieb1]

Potential solutions:

1. Instead of casting to int, cast to np.Int8 (fewer bytes, so less memory on disk)
2. Cast smp_grid before trying to multiply by mask array
3. Instead of converting smp_grid and mask to numpy arrays, convert to dask arrays and perform operation in chunks