Fix potential XSS vulnerability in break_long_headers template filter

[ch4n3-yoon]

Description

The header input is now properly escaped before splitting and joining with <br> tags. This prevents potential XSS attacks if the header contains unsanitized user input.

This pull request addresses a potential XSS vulnerability in the break_long_headers template filter. By escaping the header input before processing, the risk of XSS attacks is mitigated.

[browniebroke]

Probably deserves a test

[tomchristie]

@browniebroke Let's prioritise getting this sorted, rather than waiting on a test case.

I'd marginally prefer #9438 over this, since the line break isn't actually required however we should just go with whatever gets this resolved as quickly as possible at this point.

[ch4n3-yoon]

I've identified a potential XSS vulnerability related to the break_long_headers template filter used in the rest_framework/base.html template file by APIView. This file employs the break_long_headers template filter, making the following code vulnerable to XSS attacks due to unsanitized user input:

# views.py
from rest_framework.views import APIView
from rest_framework.response import Response

class Index(APIView):
    def get(self, request):
        username = request.GET.get('username', '')

        response = Response('OK')
        response['Location'] = f'https://x.com/{username}'
        return response

# urls.py
from django.urls import path
urlpatterns = [ path('api/', Index.as_view()), ]

I believe it is essential to register this issue as a CVE to ensure that users of earlier versions of DRF are aware and can manage this vulnerability appropriately. Your thoughts on this?