fix(fetch): add fallback for streaming SSR content and support configurable timeout

[AbhiPra24]

Summary

Fixes #3878 and #4448.

1. Fallback for Progressive/Streaming SSR Content (Fixes #3878)

• Problem: Mozilla Readability (readabilipy) strips HTML elements styled with visibility: hidden or off-screen positioning commonly used in modern SSR skeleton hydration (Next.js, Remix). Only the loading shell text survived, silently dropping up to 85% of the page.
• Fix: In extract_content_from_html(), if Readability returns empty content or extracted content that is significantly shorter than the source HTML (< len(html) * 0.05), fall back to markdownifying the source HTML directly via markdownify.

2. Configurable Timeout (Fixes #4448)

• Problem: 30-second network timeout was hardcoded in fetch_url.
• Fix: Added DEFAULT_TIMEOUT = 30.0, support for MCP_FETCH_TIMEOUT environment variable, --timeout CLI flag, and an optional timeout parameter in the Fetch tool input schema.

Verification

• Added 16 new test cases in src/fetch/tests/test_server.py covering SSR fallback, timeout resolution precedence, robots.txt timeout, and Fetch model validation.
• pytest: 36/36 passed.
• ruff check and pyright: 0 errors.

---

cc @olaservo for review when available. Thank you!