There are cases where metadata .tsv files get malformed because of uncleaned free-form text. This included sentences, sentence sources, sentences in reported and reasons where people use "other" option and write free form.

The problems were reported and analyzed in these topics:

• [BUG] Sentence input is not fully cleaned in "write", thus errors in "*_sentences.tsv" #4406
• [BUG] reported.tsv has broken rows due to LF & TAB characters in sentence and reason fields #4429
• [BUG] Malformed tables (*_sentences.tsv) for nb-NO locale #4839

The "sentence" in *_sentences.tsv were previously fixed here.

This PR extents this to:

• Source fields in *_sentences.tsv files
• Sentence and reason fields in reported.tsv

Note: This has NOT been tested on real data, but should be OK as the modifications were simple. Should be tested before the next release, if merged.