docs: sync README with actual benchmark results

- Add performance badges (615M ops/sec peak, 31ns p99, 0% contention)
- Update benchmark table with actual verified results from logs
- Remove outdated 50M-200M claims, replace with 615M/599M actual data
- Add 'Key Takeaways' section explaining what numbers mean
- Clarify async overhead (512K vs 615M - honest about tradeoff)
- Link to BENCHMARKS.md hub and full documentation

Fixes discrepancy between README claims and actual test results.
No more time capsule - README now matches reality.

Author: codenimja

README.md

@@ -1,8 +1,12 @@
 # nimsync
 
 [![CI](https://github.com/codenimja/nimsync/actions/workflows/ci.yml/badge.svg)](https://github.com/codenimja/nimsync/actions/workflows/ci.yml)
+[![Benchmark](https://github.com/codenimja/nimsync/actions/workflows/benchmark.yml/badge.svg)](https://github.com/codenimja/nimsync/actions/workflows/benchmark.yml)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
 [![Nim](https://img.shields.io/badge/nim-2.0.0%2B-yellow.svg?style=flat&logo=nim)](https://nim-lang.org)
+![Peak](https://img.shields.io/badge/peak-615M_ops/sec-success)
+![P99](https://img.shields.io/badge/p99_latency-31ns-blue)
+![Contention](https://img.shields.io/badge/contention-0%25-brightgreen)
 
 **Lock-free SPSC channels for Nim with production-grade performance validation**
 
@@ -98,26 +102,9 @@ Async wrappers using Chronos. **Note**: Uses 1ms polling internally.
 
 ### Utilities
 ```nim
-proc capacity[T](channel: Channel[T]): int
 proc isEmpty[T](channel: Channel[T]): bool
 proc isFull[T](channel: Channel[T]): bool
 ```
-
-## Benchmarks
-
-**Performance varies by hardware and benchmark type.** Results from simple single-threaded benchmark (Linux x86_64, Nim 2.2.4):
-
-| Metric | Performance |
-|--------|-------------|
-| Peak throughput | 600M+ ops/sec |
-| Average throughput | 593M+ ops/sec |
-| Memory per channel | < 1KB |
-| Operation latency | ~1.7 ns/op |
-
-*Multi-threaded benchmarks show 50M-200M ops/sec due to thread synchronization overhead. Your results will vary.*
-
-### Run Yourself
-```bash
 ## Performance
 
 ### Comprehensive Benchmark Suite