Fix asset scan result aggregation

[anpa1200]

What changed

• aggregate repeated Nmap service observations from multiple CDN/load-balancer addresses into unique protocol/port services
• retain all resolved endpoint evidence and raw per-host scan records
• deduplicate assessment findings in the API and normalize historical saved scan output in the frontend
• show affected endpoints in the service table
• add regression tests for duplicate service observations and findings

Root cause

The scanner correctly scanned each resolved address, but the API counted each per-address observation as a separate service and the frontend flattened all host rows. A hostname with four CDN addresses therefore displayed four copies of ports 80 and 443.

User impact

Threat Radar now reports the logical exposure once while preserving the underlying endpoint evidence for analyst review. Existing saved scans are also rendered without duplicate rows.

Validation

• python -m pytest -q -o addopts='' tests/unit/test_asset_vulnerability_scanner.py — 12 passed
• Ruff on changed backend files — passed
• frontend ESLint — passed
• frontend production TypeScript/Vite build — passed
• Threat Radar integration test body passed; the repository test process still hangs during global teardown