fix(fetcher): restore Stack Overflow Q&A extraction

[chaliy]

What

Fix Stack Overflow question fetching so Q&A pages return meaningful markdown again, including the question body, accepted/top answers, code snippets, links, and metadata.

Why

The Stack Exchange API rejects the old withbody_markdown filter with HTTP 400, so the specialized fetcher could no longer return the requested Q&A content.

How

• Use the supported withbody API filter for questions and answers.
• Convert Stack Exchange HTML bodies through FetchKit's existing HTML-to-markdown converter.
• Add a minimal-content guard so empty/block-like Q&A responses are not reported as successful.
• Add a mocked regression test for the reported Stack Overflow question shape.
• Add an Unreleased changelog note.

Risk

• Low
• Stack Overflow/Stack Exchange fetch output formatting changes from API-provided markdown to FetchKit-converted markdown, but this restores working content and uses the same converter as normal HTML fetches.

Security

• Reviewed SSRF/network handling: API host remains hardcoded and policy-pinned through existing transport_request flow.
• Reviewed resource exhaustion: API body reads still use existing max body size and timeout handling.
• Reviewed HTML conversion: uses existing converter that strips unsafe/noisy elements.

Checklist

• Unit tests are passed
• Smoke tests are passed
• Documentation/changelog is updated
• Specs are up to date and not in conflict
• Security review performed

Validation:

• cargo test --workspace
• cargo build --workspace --exclude fetchkit-python
• cargo run -p fetchkit-cli -- --help
• cargo run -p fetchkit --example fetch_urls
• MCP initialize smoke via cargo run -p fetchkit-cli -- mcp
• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets -- -D warnings
• RUSTDOCFLAGS="-D warnings" cargo doc --workspace --no-deps