fix: prevent ValueError in break_up_import when \r in source shifts line numbers

[koteshyelamati]

Fixes #326

Problem

fix_code() raises ValueError: not enough values to unpack (expected 2, got 1) in break_up_import when the source contains a bare \r character (e.g. a Windows CR without LF, or a CR mid-line):

import autoflake
s = """
from t import (
\r  a,
    b,
)
"""
autoflake.fix_code(s)  # ValueError: not enough values to unpack (expected 2, got 1)

Root cause

io.StringIO.readlines() does not treat \r alone as a line separator, but Python's tokenizer (and str.splitlines()) does. When pyflakes analyses the source, it uses splitlines(), so a bare \r is counted as an extra line. This causes the line numbers that pyflakes marks as import lines to be offset from the lines that filter_code sees via readlines(). As a result, break_up_import is sometimes called with a continuation line that contains no import keyword, causing the re.split() to return a 1-element list and the tuple unpacking to fail.

Fix

Add a guard in break_up_import: if re.split does not produce exactly 2 parts (i.e. no import keyword was found), return the line unchanged rather than crashing. This is a minimal, defensive fix at the point of failure.