fix: resolve absorber item crash, scrape traversal, and security ratchets

[blisspixel]

Hardens error handling and security controls across scraping, expert skills, and webhooks:

• Fix KnowledgeAbsorber.absorb AttributeError crash on string/scalar response items
• Enforce http/https schemes on LinkExtractor results, dropping javascript:, mailto:, data:, file:
• Sanitize company and project names against path traversal in scrape_for_company_research and scrape_for_documentation
• Delimit and sanitize untrusted scraped content in ContentSynthesizer before LLM prompt assembly
• Guard generator slugify against empty strings, punctuation-only names, and Windows reserved device stems
• Validate busy_timeout_ms in conversation connect_database as positive integer
• Handle TimeoutExpired and OSError in NgrokTunnel.stop
• Use canonical is_loopback_bind_host in webhook server startup check
• Mark usedforsecurity=False on non-crypto MD5 hashes and annotate parameterized SQL marker in persistence
• Tighten Bandit S ratchet baseline from 52 to 48