add flop counter for mul operation

[jixiege]

Summary: add flop counter for elementwise product between tensors.

Though there is an implementation using elementwise_flop_counter(1) in #99, but it does not support broadcasting to a common shape.

[ppwwyyxx]

Just use elementwise_flop_counter(0, 1). It supports broadcasting.

[jixiege]

Just use elementwise_flop_counter(0, 1). It supports broadcasting.

That's right, should be elementwise_flop_counter(0, 1) instead of elementwise_flop_counter(1).

Thanks for your correct.