chore: add basic benchmark results

Author: pi0

README.md

@@ -62,6 +62,60 @@ const message = capnp.Message.fromArrayBuffer(buffer);
 const struct = message.getRoot(MyStruct);
 ```
 
+## Benchmarks
+
+In theory, since capnp does not needs any serialization and deserialization, it is INFINITY TIMES faster than any other protocol as "parse" only happens on field access.
+
+To be more fair, we [benchmark](./test//benchmark) by accessing fields in different ways. It is still up to **%65** faster even on a small JSON payload than `JSON.parse`.
+
+<details>
+
+<summary>results</summary>
+
+```
+$ node --import jiti/register ./test/benchmark/bench.ts
+
+cpu: Apple M2
+runtime: node v20.15.1 (arm64-darwin)
+
+benchmark                                   time (avg)             (min … max)       p75       p99      p999
+------------------------------------------------------------------------------ -----------------------------
+• iteration over deeply nested lists
+------------------------------------------------------------------------------ -----------------------------
+capnp.Message(<buff>)                    1'357 ns/iter   (1'119 ns … 2'546 ns)  1'451 ns  1'685 ns  2'546 ns
+JSON.parse(<string>)                     1'990 ns/iter   (1'911 ns … 2'368 ns)  2'003 ns  2'150 ns  2'368 ns
+JSON.parse(TextDecoder.decode(<buff>))   2'166 ns/iter   (2'130 ns … 2'548 ns)  2'171 ns  2'544 ns  2'548 ns
+
+summary for iteration over deeply nested lists
+  capnp.Message(<buff>)
+   1.47x faster than JSON.parse(<string>)
+   1.6x faster than JSON.parse(TextDecoder.decode(<buff>))
+
+• top level list length access
+------------------------------------------------------------------------------ -----------------------------
+capnp.Message(<buff>)                    1'662 ns/iter   (1'340 ns … 2'293 ns)  1'794 ns  2'263 ns  2'293 ns
+JSON.parse(<string>)                     2'156 ns/iter   (2'100 ns … 2'620 ns)  2'157 ns  2'378 ns  2'620 ns
+JSON.parse(TextDecoder.decode(<buff>))   2'291 ns/iter   (2'240 ns … 2'492 ns)  2'311 ns  2'454 ns  2'492 ns
+
+summary for top level list length access
+  capnp.Message(<buff>)
+   1.3x faster than JSON.parse(<string>)
+   1.38x faster than JSON.parse(TextDecoder.decode(<buff>))
+
+• parse
+------------------------------------------------------------------------------ -----------------------------
+capnp.Message(<buff>).getRoot()          1'328 ns/iter   (1'110 ns … 1'710 ns)  1'456 ns  1'688 ns  1'710 ns
+JSON.parse(<string>)                     2'043 ns/iter   (1'982 ns … 2'222 ns)  2'107 ns  2'213 ns  2'222 ns
+JSON.parse(TextDecoder.decode(<buff>))   2'190 ns/iter   (2'111 ns … 2'349 ns)  2'264 ns  2'338 ns  2'349 ns
+
+summary for parse
+  capnp.Message(<buff>).getRoot()
+   1.54x faster than JSON.parse(<string>)
+   1.65x faster than JSON.parse(TextDecoder.decode(<buff>))
+```
+
+</details>
+
 ## Development
 
 <details>