updated simple matmul demo to use KernelCollection generate function rather than calling individual generator

added evaluation of lid/gid strides using kernel params, which enables more precise stride matching so that local sizes are not needed in demo global memory features

removed adjust_local_temp_var_storage, increased time trials, added random starting guess option, changed overlap demo size on stout

added rept vals to local tile sum problem size tags to enable different measurement set for AMD card

removed repetition from local tile sum knl because of weird performance (compiler may be doing some kind of loop optimization