Heuristic to determine batch_size and number of parallel jobs

[markbader]

Detailed Description

• Affected library: webknossos
• Conversion of large image stacks can lead to OOM issues. The defaults for batch_size and the number of processes used for parallelization should depend on the input image size and the available RAM. Currently the example scripts shown on our official documentation fail in cases like Image stack to dataset process can exhaust ram for large datasets #1146. Instead, the examples should work out of the box.