Secure session HKDF transcript and compatibility tests

[Anupam-USP]

This PR adds regression coverage for the secure-session migration by validating HKDF-based key derivation and compatibility behavior. It introduces tests that confirm:

• both peers derive the same session key from a canonical transcript
• changing the protocol kind, peer identity, or ephemeral keys changes the derived key
• unsupported secure-session versions fail with a clear error

The implementation now derives session keys from a canonical transcript using HKDF and enforces a secure-session version gate during the handshake, preserving compatibility expectations while protecting the migration path.

Resolves #38

[ethical-buddy]

Thanks for putting this together. The direction is good: moving secure-session key derivation to HKDF, adding transcript-based key separation, and testing version mismatch behavior are all the right areas to focus on.

We are going with PR #35 for this issue because that PR resolves the same problem more completely. In particular, PR #35 binds the transcript to the full static Ed25519 public keys instead of truncated node IDs, includes the explicit version-gated migration path, and has broader focused coverage for transcript canonicalization, key separation, public-key binding, and mixed-version behavior.

Your PR still had some good ideas. In particular, including the secure-session version in the signed handshake payload was a useful hardening detail, and we carried that idea into the final cleanup before merging PR #35.

If you want to help next, useful follow-up areas would be:

• add or improve compatibility tests around old/new secure-session peers
• help clean up the existing go test ./... failures from non-constant fmt.Errorf calls
• help fix the current Linux stable-gate dependency gap for libXxf86vm
• review the secure-session docs or add a short protocol note explaining the v2 transcript format and migration behavior

Appreciate the contribution. This PR was close, but PR #35 covers the same fix with the stronger transcript binding we need.