Introduce chunk size when reading dark and reference input

[tobiashuste]

So far, the whole files are loaded in memory --> possibly high memory consumption.
To prevent memory problems, introduce a configurable chunk size to load and average only parts of the dark and reference input data.