adjust initial caps on middleware steps

[lucix-aws]

Adjusts the initial size of each middleware step. New base value of 8 was selected based on the downstream aws-sdk-go-v2, where initialize/serialize/build typically have a little under that many middlewares. Finalize and Deserialize are typically double that size.

Before change w/ 1024*1024 mock sqs ListQueues requests:

• ~40GiB allocated
• ~415M objects allocated

After:

• ~37.5GiB allocated
• ~400M objects allocated

If you'd like to run this yourself

Run the test code (make sure you checkout this patch and modreplace in your test project):

package main

import (
	"context"
	"fmt"
	"io"
	"log"
	"net/http"
	_ "net/http/pprof"
	"strings"
	"time"

	"github.com/aws/aws-sdk-go-v2/config"
	"github.com/aws/aws-sdk-go-v2/service/sqs"
	smithyhttp "github.com/aws/smithy-go/transport/http"
)

func main() {
	cfg, err := config.LoadDefaultConfig(context.Background())
	if err != nil {
		panic(err)
	}

	svc := sqs.NewFromConfig(cfg, func(o *sqs.Options) {
		o.HTTPClient = smithyhttp.ClientDoFunc(func(r *http.Request) (*http.Response, error) {
			return &http.Response{
				StatusCode: 200,
				Body:       io.NopCloser(strings.NewReader("{}")),
			}, nil
		})
	})

	go func() {
		log.Println(http.ListenAndServe("localhost:6060", nil))
	}()

	fmt.Println("wait")
	time.Sleep(5 * time.Second)
	fmt.Println("run")

	for i := 0; i < 1024*1024; i++ {
		fmt.Println(i)
		out, err := svc.ListQueues(context.Background(), nil)
		if err != nil {
			panic(err)
		}
		_ = out
	}

	fmt.Println("ready")
	for {
	}
}

Once the "ready" is printed, run the pprof tool to get the visual:

go install github.com/google/pprof@latest # if you don't have it
pprof -http=localhost:9090 localhost:6060/debug/pprof/allocs

See aws/aws-sdk-go-v2#3197

[Madrigal]

Thanks for including the repro steps! I'm sure they'll also come handy later on