preserve_blank_lines=True counts each CRLF as two line breaks

[APITester26]

With preserve_blank_lines=True, a Windows line ending (\r\n) is counted as two line endings, so every CRLF in the input adds an extra blank line to the output.

>>> import titlecase
>>> titlecase.titlecase("Line1\r\n\r\nLine2", preserve_blank_lines=True)
'Line1\n\n\n\nLine2'   # expected: 'Line1\n\nLine2'

input	actual	expected
'a\r\nb' (no blank line)	'A\n\nB'	'A\nB'
'Line1\r\nLine2\r\nLine3'	'Line1\n\nLine2\n\nLine3'	'Line1\nLine2\nLine3'
'Line1\r\n\r\nLine2' (one blank line)	'Line1\n\n\n\nLine2'	'Line1\n\nLine2'

LF input is handled correctly: 'a\nb' gives 'A\nB' and 'a\n\nb' gives 'A\n\nB'.

The cause is in titlecase/__init__.py, in the preserve_blank_lines branch:

if preserve_blank_lines:
    lines = regex.split('[\r\n]', text)
else:
    lines = regex.split('[\r\n]+', text)

\r\n is two characters but one line ending, so split produces an empty element between the two, and joining the pieces with "\n" turns it into a blank line. The default branch (preserve_blank_lines=False) uses [\r\n]+ and is not affected.

This reaches any caller that passes CRLF text to the library API: HTTP response bodies, subprocess output, files read with newline='', output from Windows tools. The CLI is not affected, because it opens files in text mode and universal newlines convert CRLF to LF before titlecase sees the text.

A fix would be to treat one line ending as one separator, for example:

lines = regex.split(r'\r\n|\r|\n', text)

TestBlankLines.test_complex_blanks only uses \n, so this path currently has no coverage.

The same behaviour is present in the 2.4.1 release, so it is not specific to main.