update bench

[zhongwencool]

Update:

1. Using list_to_binary/1 replace io:iodata_to_binary/1, because we already known it's a list.
2. Update << random, acc::binary>> replace << acc::binary, random>>, make sure benchmark more fair.

Great to know!!
[item|list] is faster than <<item, list/binary>>
But binary construction is more flexible then list construction
we can do

item = 100,
list = <<1,2,,3,4,5,6>>
<<Item, list/binary>> # add item at head
or 
<<list/binary, item>> # add item at tail

if it a list, we only can do

[item|list] # add item at head, can't add it to tail.

[secretworry]

Cool, Being more flexible has its downside too, I guess.

[secretworry]

Extending a binary involves malloc&copy, while prepending a list would just incur assigning a pointer, I think that's the reason