refactor: structured body (array|object, never string) for canonicalization

Message bodies are always arrays of strings (text lines) or objects
(structured data like PUBKEY). Never raw strings. This enables:
- Multiline messages without escape sequences
- Deterministic JSON canonicalization (RFC 8785 JCS) for signing
- Structured data where needed

Update all schemas: body fields use array type with string items.
Update message.json envelope: body is oneOf[array, object], id is UUID.
Update README: message envelope table, examples, and canonicalization docs.
Update schema/README.md: field types, examples with array bodies.

schema/commands/PONG.json

@@ -6,17 +6,16 @@
   "$ref": "../message.json",
   "properties": {
     "command": { "const": "PONG" },
-    "from": {
-      "type": "string",
-      "description": "Responding server name."
-    },
+    "from": { "type": "string", "description": "Responding server name." },
     "body": {
-      "type": "string",
-      "description": "Echoed token from PING."
+      "type": "array",
+      "items": { "type": "string" },
+      "description": "Echoed token from PING (single-element array).",
+      "maxItems": 1
     }
   },
   "required": ["command"],
   "examples": [
-    { "command": "PONG", "from": "server.example.com", "body": "1707580000" }
+    { "command": "PONG", "from": "server.example.com", "body": ["1707580000"] }
   ]
 }