Optimize execution trace generation for Memory co-processor

[grjte]

Currently, when the execution trace is generated by the Memory co-processor, we loop through the trace and compute the inverse of the delta value at each iteration (the code is here). Instead, we should switch to using batch inversion.