As a general rule of physics: for a strictly linear system, it matters not: half an hour at 150 W followed by half an hour's break is the same as an hour at 75 W. But nothing in life is ever linear, and bicycles and bicyclists are certainly not. Wind resistance increases faster than linear, and assuming you achieve a higher speed in part by pushing harder rather than spinning quicker, I think that is non-linear too. In the presence of non-linearities, the lowest-energy solution is always the one with least variation.
So I suggest that for a given average speed over a given distance, the lowest energy consumtion will come from doing it as uniformly as possible. Doing part of it quicker then stopping will, I suggest, take more energy. (Whether "as uniformly as possible" means uniform speed or uniform power output, or put another way, what is the optimum balance between those two, will depend on how much of your energy is expended on the road and how much in wasted heat in your muscles.)
Of course, I've made the naive physicist's assumption that energy used = weight lost - any physiologists out there are welcome to explain why that may not be true...