Feature request: Allow regex flags for `str.split`

[tsoernes]

Most string functions allow case and flags for matching; str.split does not.

To add to this, I think the API should be updated to more closely align with str.replace.
The current logic for determining whether the split pattern is a string or a regex is not clear from the docs (#25549), and does not match Python's built-in str.split. Looking at the code, the current (undocumented) logic is:

• If the pattern is None, or has length 1, it's a string split.
• Else it's a regex split.

However, the Python built-in str.split function accepts string patterns with length greater than 1, and doesn't accept regex patterns at all. Updating the function to include a regex=True/False argument would make this explicit, and allow string splits on delimiters with length greater than 1.

[stacymiller]

A side note: as currently Pandas str.split accepts compiled regexes, all flags can be passed to re.compile():

Without flags:

>>> pd.Series(['first line\nsecond line', 'third line']).str.split(
...     re.compile('$'), regex=True, expand=True
... )
                         0 1
0  first line\nsecond line  
1               third line  

With flags:

>>> pd.Series(['first line\nsecond line', 'third line']).str.split(
...     re.compile('$', flags=re.MULTILINE), regex=True, expand=True
... )
            0              1     2
0  first line  \nsecond line      
1  third line                 None