feat: add DIG_MESSAGE opcode const (220) for directed dig-message envelopes

[MichaelTaylor3d]

Task

Additive canonical opcode const pub const DIG_MESSAGE: u8 = 220 for the directed dig-message envelope. Part of WU6 (epic #796) — the release-first sub-step exposing the canonical opcode to the dig-message adoption consumers.

What changed

• DIG_MESSAGE = 220 — first opcode of the free 220..=255 band (the 200..=219 consensus band, DigMessageType, is untouched — §5.1 additive, no renumber).
• Doc: the envelope rides as OPAQUE bytes in DigMessage.data; the transport never seals/opens it (end-to-end sealing is dig-message's job, §5.4).
• 2 tests: opcode frames a real DigMessage + round-trips with msg_type intact; 220 is not a consensus DigMessageType.

Blast radius

Purely additive public API — no existing symbol touched. Consumers reference dig_protocol::DIG_MESSAGE (dig-gossip mirrors it as ProtocolMessageTypes::DigMessage + dig_gossip::DIG_MESSAGE; dig-relay-protocol #874, dig-ipc-protocol #1074, ...).

Version

MINOR bump 0.1.2 → 0.2.0 (new public API).

Refs #1164 #796

🤖 Generated with Claude Code

[MichaelTaylor3d]

PASS (loop-reviewer, fresh context).

Verified against the WU6/#1164 acceptance criteria:

• Additive-only — pub const DIG_MESSAGE: u8 = 220 is the first opcode of the free 220..=255 band; the 200..=219 consensus band (DigMessageType, RegisterPeer=218/RegisterAck=219) is untouched. No renumber/repurpose of any existing type (§5.1 satisfied). Doc-comment is intent-revealing and states the OPAQUE-envelope + e2e-sealing-is-dig-message's-job contract (§5.4).
• Tests — dig_message_opcode_frames_and_round_trips exercises 220 through the real encoder with msg_type intact; dig_message_opcode_is_not_a_consensus_type proves 220 is not a DigMessageType discriminant (bands can't collide). Coverage gate green.
• Version — 0.1.2 → 0.2.0 MINOR is correct for a new public API (§2.4).
• Gates — fmt/clippy/test/docs/coverage/commitlint/version-increment all green; CodeQL clean.
• dig-constants check — protocol opcodes are canonically homed in dig-protocol (not dig-constants, which holds asset-ids/addresses/ports); this crate IS the canonical home #874/#1074 consumers reference. Correct placement.

Trivial + correct. No findings. Merge-order note: land this then publish, or land dig-gossip#7 first (it is self-contained) — either is safe.

(Recorded as --comment, not --approve: self-authored PR, 422 on self-approval; branch protection gates on green checks + zero unresolved threads, both satisfied.)