tokenize.detect_encoding has issues with \r

Tokenize does not correctly detect encodings when the carriage return character is involved. In this program, it does not detect the header, since it does not consider the \r to start a new line.

import tokenize
import runpy

code = """\r# coding: big5

print('錯誤的答案')
"""

with open('file.py', 'w', encoding='big5') as file:
    file.write(code)
with open('file.py', 'rb') as file:
    print(tokenize.detect_encoding(file.readline))
    # result: ('utf-8', [b'\r# coding: big5\r\n', b'\r\n'])

runpy.run_path('file.py')
# prints 錯誤的答案 as required

The issue names tokenize.detect_encoding and provides a complete carriage-return reproduction. Start by running the supplied script, then locate the encoding-detection implementation and existing tests; done means the coding header is detected as big5 when \r precedes it, with regression coverage for this input.