Investigate replacing recursive preprocessing algorithm with sequential one

[danielytics]

The chance of overflowing the stack for large templates should be reduced. Initial tests have also shown sequential algorithm to be a little bit faster. Need to investigate if this is something worth implementing properly.