Source Detection is restructured

[kenfus]

Currently, Source Detection is quite wierd:

• you call detected sources in image on a class called SourceDetectionResult, which returns itself, but this time with the detected sources as an attribute.
• Usage of cls makes maybe the code more clean but overall less readable.
• Thanks to parallelization, there is now a class, which inherits from a class, which inherits from a class and one of them returns itself with cls, but actually you call a function, which this class does not have but inherits this from another class.

Thing should be split:

• One class called SourceDetection, which results SourceDetectionResult with the results.

Other improvements:

• No intermediate files are being kept, if they are not needed (such as source.csv).
• Clean up better after source detection.