// Check the cache again because singleflight can only dedup calls
		// that overlap concurrently. It's possible for 2 concurrent
		// requests to miss the cache, resulting in 2 load() calls. An
		// unfortunate goroutine scheduling would result in this callback
		// being run twice, serially. If we don't check the cache again,
		// cache.nbytes would be incremented below even though there will
		// be only one entry for this key.
		//
		// Consider the following serialized event ordering for two
		// goroutines in which this callback gets called twice for the
		// same key:
		// 1: Get("key")
		// 2: Get("key")
		// 1: lookupCache("key")
		// 2: lookupCache("key")
		// 1: load("key")
		// 2: load("key")
		// 1: loadGroup.Do("key", fn)
		// 1: fn()
		// 2: loadGroup.Do("key", fn)
		// 2: fn()

fn() {
 再次查询缓存是否已加载,已加载直接返回
 执行 fn 的逻辑
}