feat: add OpenRouter LLM chat plugin (!ask, !chat)

Single-shot (!ask) and conversational (!chat) LLM commands backed by
OpenRouter's API. Per-user history (20 msg cap), 5s cooldown, reasoning
model fallback, and model switching via subcommands.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs/USAGE.md

@@ -187,6 +187,8 @@ unchanged. The server name is derived from the hostname automatically.
 | `!username list` | Show available services by category |
 | `!alert <add\|del\|list\|check\|info\|history>` | Keyword alert subscriptions across platforms |
 | `!searx <query>` | Search SearXNG and show top results |
+| `!ask <question>` | Single-shot LLM question via OpenRouter |
+| `!chat <msg\|clear\|model\|models>` | Conversational LLM chat with history |
 | `!jwt <token>` | Decode JWT header, claims, and flag issues |
 | `!mac <address\|random\|update>` | MAC OUI vendor lookup / random MAC |
 | `!abuse <ip> [ip2 ...]` | AbuseIPDB reputation check |
@@ -815,6 +817,55 @@ Title Two -- https://example.com/page2
 Title Three -- https://example.com/page3
 ```
 
+### `!ask` / `!chat` -- LLM Chat (OpenRouter)
+
+Chat with large language models via [OpenRouter](https://openrouter.ai/)'s
+API. `!ask` is stateless (single question), `!chat` maintains per-user
+conversation history.
+
+```
+!ask <question>              Single-shot question (no history)
+!chat <message>              Chat with conversation history
+!chat clear                  Clear your history
+!chat model                  Show current model
+!chat model <name>           Switch model
+!chat models                 List suggested free models
+```
+
+Output format:
+
+```
+<alice> !ask what is DNS
+<derp> DNS (Domain Name System) translates domain names to IP addresses...
+
+<alice> !chat explain TCP
+<derp> TCP is a connection-oriented transport protocol...
+<alice> !chat how does the handshake work
+<derp> The TCP three-way handshake: SYN, SYN-ACK, ACK...
+```
+
+- Open to all users, works in channels and PMs
+- Per-user cooldown: 5 seconds between requests
+- Conversation history capped at 20 messages per user (ephemeral, not
+  persisted across restarts)
+- Responses truncated to 400 characters; multi-line replies use paste overflow
+- Default model: `openrouter/auto` (auto-routes to best available free model)
+- Reasoning models (DeepSeek R1) are handled transparently -- falls back to
+  the `reasoning` field when `content` is empty
+- Rate limit errors (HTTP 429) produce a clear user-facing message
+
+Configuration:
+
+```toml
+[openrouter]
+api_key = ""                  # or set OPENROUTER_API_KEY env var
+model = "openrouter/auto"     # default model
+system_prompt = "You are a helpful IRC bot assistant. Keep responses concise and under 200 words."
+```
+
+API key: set `OPENROUTER_API_KEY` env var (preferred) or `api_key` under
+`[openrouter]` in config. The env var takes precedence.
+
 ### `!alert` -- Keyword Alert Subscriptions
 
 Search keywords across 27 platforms and announce new results. Unlike