/claim #286

Summary

Fixes #286

Implements a multi-provider LLM routing system inspired by LiteLLM for the EdgeChains JS SDK.

Features

1. Load Balancing — round-robin, least-tokens, latency-based, cost-based strategies with RPM/TPM rate limiting
2. Streaming — AsyncGenerator-based with automatic fallback on stream errors
3. Token Usage Tracking — per-model counters with getUsage() and resetUsageCounters()
4. Logging — Sentry breadcrumbs + error capture, PostHog event tracking, custom callbacks

Testing

24 tests (unit + e2e with mock servers), all passing. TypeScript compiles clean.