MCP: v1 layout wire encoding

[xinyuan-dev]

This PR implements the encoding and decoding of logical Chunk into/from v1 wire format (Bytes).

The concept of PacketLayout gets absorbed into wire/v1.rs and EncodingScheme. Now v1 packet layout is simplified into two parts: header (sig, signed bytes) and body (merkle proof, chunk header, symbol).

Clarified separation of responsibility:

• wire/v1.rs owns the knowledge of packet format (segment length, positions of fields), decoding & encoding of the bytes
• EncodingScheme is logical - it owns the knowledge of the app_msg_len, the depth and their calculation as well as other header fields.

---

Changes:

EncodingScheme for D25 now includes three fields: app_msg_len, depth, coarse_ts instead of two: app_msg_len, unix_ts. The extra depth field makes validation cheaper and obviate the need to pre-compute packet layout. The coarse-grained ts is better at expressing the semantic of equality on EncodingScheme.

The stub symbol codec implementation is changed to a repetition code to accommodate for the enforced symbol length.

Another conceptual simplification is the delayed verification of a chunk:

• essential v1 packet well-formedness check: done at parse time
• proposal header check (slot valid, author is proposer, depth canonical): done in SlotRaptorcast (aka authenticate, shared with chorus for proposal headers received in consensus message)
• per-chunk merkle proof verification: done in RaptorcastInstance; so a chunk with invalid merkle proof still gets its header registered for potential equivocation.