Potentially losing comment state between reads on CommentRemovingReader

In more extensive testing on real world data I found a case that was not properly handled in the rewrite towards BufferedReader. If. a read ends between the first / and the subsequent / or * for comment start, the detection will currently fail.

To fix this I refactored the detection logic that should now also be significantly better readable always capturing the previous char on class level allowing to check for char double char sequences that may be split by the reads.