refactor(framing): replace Sv2Frame dual-Option fields with FrameBody enum

[ThomsenDrake]

Fixes #2111, #2112, #2113

Summary

Replaces Sv2Frame<T, B>'s dual Option<T> + Option<B> fields with a type-safe FrameBody<T, B> enum (Payload(T) | Serialized(B)). This makes the invariant that exactly one of payload/serialized is present impossible to violate by construction, eliminating 3 panic!("Impossible state") branches.

Changes

Core refactor (fixes #2111)

• Added FrameBody<T, B> enum with Payload(T) and Serialized(B) variants
• Sv2Frame<T, B> now uses a single body: FrameBody<T, B> field
• serialize(), encoded_length() are now exhaustive matches (no panic branches)
• map() updated to match on body variants

API soundness (fixes #2112)

• Sv2Frame::get_header() now returns Header directly (was Option<Header> — always returned Some)
• Updated all call sites (codec-sv2 decoder tests, examples)

API soundness (fixes #2113)

• HandShakeFrame::from_bytes() now returns Self directly (was Result<Self, isize> — always returned Ok)

Testing

All 9 tests pass:

• framing-sv2: 3/3 (default)
• codec-sv2: 6/6 (default + with_buffer_pool features)

[Alkamal01]

Reviewed the changes and verified them by pulling the branch locally.

The refactor to FrameBody is a great improvement — it eliminates the dual-Option field logic and correctly removes the panic! branches in favors of exhaustive matching. The return type refinements for get_header and HandShakeFrame::from_bytes also help clean up the public API.

All tests for framing-sv2 and codec-sv2 (unencrypted, noise-sv2, and with_buffer_pool) passed on my side. Nice work!

[ThomsenDrake]

Thanks for taking the time to verify locally, @Alkamal01! Appreciate the thorough review — glad the exhaustive matching and API cleanups checked out across all test suites.

[Alkamal01]

Thank you! Happy to help👍

[Shourya742]

Thanks for taking this up. We are yet to triage on how to move forward on protocol crates refactoring, as it has ripple effects on all the downstream crates.