ocrd-sbb-textline-detector uses ocrd_page's parse() to parse XML input,
which writes the XML to stdout by default.
Fix this by silencing it using parse()'s silence=True.
ocrd_sbb_textline_detection used the output XML by main.py as is, and
– by doing this – threw away any input data from the input PAGE,
including the critical pc:AlternativeImage and the less important
pc:MetadataItem.
Fix this by merging the segmentation results into a file created from
the input file.
Also add a pc:MetadataItem processingStep about the segmentation
operation.