test: DNSPolicy load-balancing strategy conflict

[averevki]

Description

• Adds test coverage for DNSPolicy load-balancing strategy conflicts in multicluster environments
• Validates that when two DNSPolicies target the same hostname with different load-balancing strategies, the second policy is rejected with proper error messaging
• Refactors dns_server and dns_server2 fixtures to be available at the multicluster package level for better reusability

Closes #789

Changes

New Features

• New test file: testsuite/tests/multicluster/test_strategy_conflict.py
  • Implements test_lb_strategy_conflict() to verify DNSPolicy strategy conflict detection
  • Tests that conflicting DNSPolicy shows Enforced=False condition
  • Validates DNS record condition shows ProviderError with CNAME/A record type conflict message
  • Verifies DNS resolution returns only the first gateway's IP address

Refactoring

• Moved dns_server fixture from testsuite/tests/multicluster/load_balanced/conftest.py to testsuite/tests/multicluster/conftest.py for package-level availability
• Simplified dns_server2 fixture in testsuite/tests/multicluster/load_balanced/conftest.py to validate and return only the second DNS server configuration
• Removed dns_config fixture from load_balanced tests (no longer needed with individual server fixtures)

Verification steps

Run the new test:

poetry run pytest -vv testsuite/tests/multicluster/test_strategy_conflict.py

# Verify existing load-balanced tests still pass:
poetry run pytest -vv testsuite/tests/multicluster/load_balanced/

[crstrn13]

LGTM! 🎉

[emmaaroche]

New test is passing, however I ran poetry run pytest -vv testsuite/tests/multicluster/load_balanced/ a few times on kua-420 pairs and kua-hub pairs and got same result:

FAILED testsuite/tests/multicluster/load_balanced/test_load_balanced_geo.py::test_load_balanced_geo - httpx.ConnectTimeout: timed out

But this is likely environmental issue, is this happening for you @crstrn13 @averevki

[azgabur]

Non-blocking, otherwise LGTM

[averevki]

@emmaaroche It is indeed an unrelated environmental issue we have, you can also check that it was failing on nightly runs for the last week. Looking into this but test is passing on hub cluster for me today