refactor: complete backend rewrite for multi-protocol proxy

Major rewrite of the Go backend to support:
- Three API format imports: openai, anthropic, compatible
- Three protocol conversions: Chat Completions, Responses, Messages
- Hub-and-spoke architecture with Chat as intermediate format

Deleted:
- opencat.go (old entry)
- store/, team/, pkg/team/, pkg/store/ (old data layer)
- internal/model/, internal/consts/ (old types)
- internal/service/team/, internal/controller/team/ (old handlers)
- llm/ (removed LLM client library, pure proxy mode)
- dist/, assets/ (old build artifacts)

Added:
- internal/store/ — 9 GORM models + multi-DB support
- internal/pkg/ — crypto (AES-GCM), apikey, jwt, ratelimit, resp, tokenizer
- internal/channel/ — channel selection, weighted LB, health checks
- internal/proxy/convert/ — 6 protocol conversion functions + SSE streaming
- internal/proxy/ — gateway with request dispatch and upstream selection
- internal/usage/ — async usage recorder with batch writes
- internal/api/ — management API (auth, users, keys, channels, models)
- Makefile for build/test/deploy

Fixed API to match frontend expectations:
- Login response wraps token in { data: { token } }
- GET /api/profile route added
- Profile response wraps user in { code, data }
- Role returned as number (10=admin, 1=user)

frontend/index.html

@@ -4,7 +4,7 @@
     <meta charset="UTF-8" />
     <link rel="icon" type="image/svg+xml" href="/src/assets/logo.svg" />
     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-    <meta name="description" content="OpenTeam — an open-source, team-shared service compatible with the OpenAI API." />
+    <meta name="description" content="OpenTeam — one endpoint for every LLM provider. Route OpenAI-compatible clients to Claude, Gemini, DeepSeek and more through a single, team-shared endpoint." />
     <meta name="theme-color" content="#ffffff" />
     <title>OpenTeam</title>
     <script>