Optimized Vandermonde matrix generation for better performance

[synapticarbors]

This performance patch addresses https://github.com/weijianzhang/MatrixDepot.jl/issues/2. Thanks to @stevengj for suggesting this version of the implementation based on discussions on Julia-User surrounding the following IJulia notebook:

http://nbviewer.ipython.org/gist/synapticarbors/26910166ab775c04c47b

[weijianzhg]

Thanks