ML-based charset encoding detector. MojibusterEncodingDetector runs structural detectors for UTF-32 / UTF-16 / UTF-8 and then a 33-class byte-bigram Naive Bayes classifier covering CJK multi-byte, EBCDIC, DOS OEM, Cyrillic, Windows single-byte, ISO-8859, and Mac encodings. This detector replaces ICU4J and juniversalchardet as the default statistical encoding detector in Tika.