missleading benchmark results

[KingOfBaboon]

Hi, it's not a real PR but some sample code to demonstrate my idea. First, thanks for this great project.

I'm trying to include this lib in my project and I was doing some benchmarking comparing native map vs bigcache get for bytes. I was getting 11ns/op for native map and getting 101ns/op for bigcache on a 4,000,000 string keys. It's 10x slower than native map.

Which is pretty off from your benchmarking result:

BenchmarkMapGetForBytes-8                    	23130621	       207.2 ns/op	      23 B/op	       1 allocs/op
BenchmarkBigCacheGetForBytes-8               12868870	       346.8 ns/op	     151 B/op	       3 allocs/op

your results indicate 1.6x slower than native map.

looking into your benchmark code, I realize this line id := rand.Intn(maxEntryCount) actually takes most of the time in the benchmark loop. So the current results are misleading to some degree.

it's not easy to exclude certain code in the loop with timers, since the timer function is also slow compared to fast operations like map lookup. I included a baseline test to give us the time spent on the rand.Intn. There should be better ways to narrow down the timer to focus on our target code.

try to run the sample code and you'll know what I mean.