Refactor breaker to link call to result

[costela]

This PR lays the groundwork e.g. for an optional prometheus wrapper, linking each execution to its result.

This refactor is a breaking change. The API now looks like:

h := hoglet.NewCircuit(
    func(ctx context.Context, foo Foo) (Bar, error) {
        // do stuff
        return Bar{}, nil
    },
    hoglet.NewSlidingWindowBreaker(10, 0.1),
    hoglet.WithFailureCondition(hoglet.IgnoreContextCancelation),
)
f, err := h.Call(context.Background(), Foo{/* ... */})
//...

Benchmarks against latest hystrix:

cpu: 12th Gen Intel(R) Core(TM) i7-1250U
BenchmarkHoglet
BenchmarkHoglet-12     	30580755	       343.3 ns/op	     233 B/op	       6 allocs/op
BenchmarkHystrix
BenchmarkHystrix-12    	13177122	       970.2 ns/op	    1217 B/op	      23 allocs/op

benchmark code

package bench_test

import (
	"context"
	"testing"

	"github.com/exaring/hoglet"
	"github.com/exaring/hystrix-go/hystrix"
)

func BenchmarkHoglet(b *testing.B) {
	circuit := hoglet.NewCircuit(
		func(context.Context, any) (any, error) { return nil, nil },
		hoglet.NewEWMABreaker(10, 0.9),
	)

	ctx := context.Background()

	b.ReportAllocs()
	b.ResetTimer()

	b.RunParallel(func(pb *testing.PB) {
		for pb.Next() {
			_, _ = circuit.Call(ctx, nil)
		}
	})
}

func BenchmarkHystrix(b *testing.B) {
	cmdName := "foo"

	hystrix.ConfigureCommand(cmdName, hystrix.CommandConfig{
		Timeout:                250,
		MaxConcurrentRequests:  0, // unlimited; otherwise we skew the comparison
		ErrorPercentThreshold:  10,
		RequestVolumeThreshold: 20,
		SleepWindow:            2000,
	})

	ctx := context.Background()

	b.ReportAllocs()
	b.ResetTimer()

	b.RunParallel(func(pb *testing.PB) {
		for pb.Next() {
			_ = hystrix.DoC(ctx, "foo", func(context.Context) (error) { return nil }, nil)
		}
	})
}

[obitech]

LGTM