fix(fetch): make Readability.js extraction opt-in

[he-yufeng]

Summary

• default mcp-server-fetch HTML extraction to readabilipy's Python parser instead of always enabling Readability.js
• add --use-readability for operators who explicitly want the Node-backed Readability.js path
• document the new opt-in and cover both default and opt-in paths in tests

Closes #4199.

To verify

• uv run pytest -q --basetemp .tmp\pytest -p no:cacheprovider
• uv run ruff check src tests
• uv run pyright src tests
• uv run python -m py_compile src\mcp_server_fetch\server.py src\mcp_server_fetch\__init__.py

[he-yufeng]

Quick follow-up: this still has the fetch package tests/build green and is scoped to making Readability.js extraction opt-in instead of changing plain fetch behavior by default.

Could a maintainer take a look when convenient?

[LuuOW]

Technical audit: Verified MCP server implementation for consistency with current SDK patterns.

[he-yufeng]

Folding this into #4281, which covers the same Readability.js-vs-Python issue (#4199). I ended up preferring an automatic fallback when Node isn't available over flipping the default, since it's a smaller behavior change for existing users. Keeping that PR and closing this one to avoid three competing approaches on the same files.