fix(plugin): bind adapter to real Hermes gateway API; add contract-test gate

- Resolve the Hermes gateway API lazily via get_adapter_class() so the adapter
  always inherits the real gateway.platforms.base.BasePlatformAdapter instead
  of the ImportError fallback stub (root cause of the AttributeError:
  'set_message_handler' crash at gateway startup).
- Pass Platform('meshtastic') enum to super().__init__ and align
  connect(*, is_reconnect=False) -> bool with the gateway contract.
- Split pure formatting utils into formatting.py; remove the stub fallback.
- Add tests/contract (L0-L3): real-base binding, ABC instantiation, gateway
  run.py wiring replay, signature and event-shape conformance, run against
  hermes-agent pinned to the deployed image source (v2026.8.31).
- Add scripts/contract-test.sh and a contract-test CI job; release publish
  now requires contract tests to pass.
- Untrack stray __pycache__ artifacts.

plugin/hermes_meshtastic/formatting.py

@@ -0,0 +1,61 @@
+"""Pure formatting utilities for LoRa-friendly text.
+
+These helpers have NO dependency on the Hermes gateway package so they can be
+imported and unit-tested in isolation (CI runs them without hermes-agent).
+"""
+
+import re
+
+MAX_LORA_MESSAGE_LENGTH = 180  # Characters per packet to prevent LoRa airtime congestion
+
+_MARKDOWN_RULES = (
+    (r"\*\*(.+?)\*\*", r"\1"),  # bold
+    (r"__(.+?)__", r"\1"),
+    (r"\*(.+?)\*", r"\1"),      # italic
+    (r"(?<!\w)_(.+?)_(?!\w)", r"\1"),
+    (r"`(.+?)`", r"\1"),        # inline code
+    (r"```[\w]*\n?", ""),       # code fences
+    (r"!\[([^\]]*)\]\(([^)]+)\)", r"\1"),
+    (r"\[([^\]]+)\]\(([^)]+)\)", r"\1 (\2)"),
+    (r"^#+\s*", ""),            # headings
+    (r"^\s*[-*+]\s+", "- "),    # normalize lists
+)
+
+
+def strip_markdown(text: str) -> str:
+    """Convert markdown formatting into plain readable text suitable for radio."""
+    if not text:
+        return ""
+    result = text
+    for pattern, replacement in _MARKDOWN_RULES:
+        result = re.sub(pattern, replacement, result, flags=re.MULTILINE)
+    # Collapse multiple blank lines
+    result = re.sub(r"\n{3,}", "\n\n", result)
+    return result.strip()
+
+
+def chunk_text(text: str, limit: int = MAX_LORA_MESSAGE_LENGTH) -> list:
+    """Split text into smaller chunks for LoRa transmission, breaking at whitespace/newlines."""
+    if not text:
+        return []
+    if len(text) <= limit:
+        return [text]
+
+    chunks = []
+    remaining = text
+    while remaining:
+        if len(remaining) <= limit:
+            chunks.append(remaining.strip())
+            break
+        split_at = remaining.rfind(" ", 0, limit)
+        if split_at <= 0 or split_at < (limit // 3):
+            newline_at = remaining.rfind("\n", 0, limit)
+            if newline_at > 0:
+                split_at = newline_at
+            else:
+                split_at = limit
+        chunk = remaining[:split_at].strip()
+        if chunk:
+            chunks.append(chunk)
+        remaining = remaining[split_at:].lstrip()
+    return chunks