Don't retry when context cancelled.

[lkysow]

Previously, when the context was cancelled we would enter the error
handling clause and the loop would be re-executed. Then the
backoff.Retry would exit immediately because the context was cancelled
and we'd run in a tight loop until the program was killed.

With this change, when we exit the backoff.Retry with an error, we check
if the context is done. If so, we exit immediately.

This functionality already existed in syncer.go:

		// Get all services with tags.
		var serviceMap map[string][]string
		var meta *api.QueryMeta
		err := backoff.Retry(func() error {
			var err error
			serviceMap, meta, err = s.Client.Catalog().Services(opts)
			return err
		}, backoff.WithContext(backoff.NewExponentialBackOff(), ctx))

		// If the context is ended, then we end
		if ctx.Err() != nil {
			return
		}

[lkysow]

Previously, this continue would have us immediately restarting the for loop. Instead we should fall down into the select statement where we retry after a defined period of time

[ishustava]

Looks good! I added a question inline, but nothing that should block this. You've mentioned that this is a bug; is it a bug that we could add tests for?

[lkysow]

Looks good! I added a question inline, but nothing that should block this. You've mentioned that this is a bug; is it a bug that we could add tests for?

It's more of a bug that we see during our unit tests because we're stopping the app many times and so it continues to loop.

Maybe we can add a test, let me think about it.

[lkysow]

@ishustava I added a test

[ishustava]

Awesome, Luke! Thanks for adding the test!