Reads all strict UTF-8 lines from one snapshot, within the encoded file-size bound. LF, CRLF and CR are removed. Empty files have no lines; a final delimiter adds no extra line.