UI's ConfigProviderWithRefresh enters tight loop after Close is called

[bpalmer-lyft]

Describe the bug

In server/server/config/config_provider_with_refresh.go , refreshConfig enters a tight loop
after the Close() method is called, because break only breaks out of select, rather than the entire for loop. If it used return instead, or a labelled break, it would exit immediately.

That is, in refreshConfig,

	for {
		select {
		case <-s.stop:
			break
		case <-s.ticker.C:
		}

		newConfig, err := s.provider.GetConfig()
                 // ...

should probably instead be

	for {
		select {
		case <-s.stop:
			return
		case <-s.ticker.C:
		}

		newConfig, err := s.provider.GetConfig()
                 // ...

To Reproduce
The bug can be seen in a test similar to

type TestConfigProvider struct {
	callCount int
}

func (tcp *TestConfigProvider) GetConfig() (*Config, error) {
	tcp.callCount++
	return &Config{
		RefreshInterval: time.Second,
	}, nil
}

func TestStopping(t *testing.T) {
	provider := new(TestConfigProvider)
	start := time.Now()
	refresh, err := NewConfigProviderWithRefresh(provider)
	if err != nil {
		t.Fatalf("unexpected error: %s", err)
	}
	if provider.callCount != 1 {
		t.Fatalf("Expected refresh to happen once but it occurred %d times", provider.callCount)
	}
	refresh.Close()
	seconds := time.Since(start).Seconds()
	if provider.callCount != 1+int(seconds) {
		t.Fatalf("Expected refresh to happen %d times but it occurred %d times", 1+int(seconds), provider.callCount)
	}
}

(You'll also see a lot of loaded new UI server configuration being output to the log before go test finishes, showing the tight loop it's entered).