Consider the following file:
```
$ cat test.txt
Thsi line contains a typo
While this line is correct
$
```

As expected, using --write-changes fixes the typo in the file:
```
$ codespell --write-changes test.txt
FIXED: test.txt
$ cat test.txt
This line contains a typo
While this line is correct
$
```

However, if we use --ignore-multiline-regex as well, we get instead:
```
$ codespell \
    --ignore-multiline-regex 'codespell:ignore-begin.*codespell:ignore-end' \
    -w \
    test.txt
FIXED: test.txt
$ cat test.txt
This line contains a typoWhile this line is correct$
```

Fix this in the self.ignore_multiline_regex != None case in
FileOpener.get_lines, by using str.splitlines instead of str.split.

Fixes:
- codespell-project#3642

No functional changes.

This makes parse_file a bit more readable, and makes the following
patch smaller.

Consider the following file:
```
$ cat test.txt
codespell:ignore-begin
Thsi line contains a typo
codespell:ignore-end
Thsi line also contains a typo
While this line is correct
```

If we use codespell to fix the second typo:
```
$ codespell \
    --ignore-multiline-regex 'codespell:ignore-begin.*codespell:ignore-end' \
    --write-changes \
    test.txt
FIXED: test.txt
```
indeed that typo is fixed, but the text matching the multiline regexp is gone:
```
$ cat test.txt



This line also contains a typo
While this line is correct
$
...

The problem is that FileOpener.get_lines (returning a list of strings)
implements --ignore-multiline-regex by blanking out the text matching the
regexp.

Fix this by changing FileOpener.get_lines to return a list of fragments, each
modeled as a tuple (ignored: bool, line_number: int, lines: list[str]), and
handling this new format elsewhere.