perf: use insertion sort algorithm

[tsctx]

Benchmark Script

import { bench, run } from 'mitata'
import { Headers } from '../lib/fetch/headers.js'
import symbols from '../lib/core/symbols.js'

const headers = new Headers(
  [
    'Sec-CH-Prefers-Color-Scheme',
    'Keep-Alive',
    'TE',
    'Sec-CH-UA-Arch',
    'Max-Forwards',
    'DPR',
    'Expect',
    'Tk',
    'DNT',
    'Link',
    'Timing-Allow-Origin',
    'Device-Memory',
    'Alt-Svc',
    'Authorization'
  ].map((v) => [v, ''])
)

const headersList = headers[symbols.kHeadersList]

const kHeadersSortedMap = Reflect.ownKeys(headersList).find(
  (c) => String(c) === 'Symbol(headers map sorted)'
)

bench('Headers@@iterator', () => {
  headersList[kHeadersSortedMap] = null
  return [...headers]
})

await run()

• main

benchmark              time (avg)             (min … max)       p75       p99      p999       
--------------------------------------------------------- -----------------------------       
Headers@@iterator   8,767 ns/iter   (5,300 ns … 3,360 µs)  6,600 ns   23.5 µs  234.8 µs

• this patch

benchmark              time (avg)             (min … max)       p75       p99      p999       
--------------------------------------------------------- -----------------------------       
Headers@@iterator   7,671 ns/iter   (4,800 ns … 2,735 µs)  5,900 ns   21.1 µs   95.2 µs