Remove upper and lower bound on BUDDY_DIP_OPT_STRIP_MINING

[meshtag]

Is your feature request related to a problem? Please describe.
As of now, there is an upper and lower bound on stride size while using dip.corr_2d where upper bound is image width and lower bound is kernel width.

Describe the solution you'd like
I wish to eliminate the above mentioned bottleneck so that dip.corr_2d can be truly independent of any image/kernel.

Describe alternatives you've considered
I think we can develop support for a novel type of extrapolation (Refer this doc for context on type of extrapolation) which can fill corresponding values related to Upper/Mid/Lower left col, Upper/Mid/Lower mid col, Upper/Mid/Lower right col in a single if block.

Additional context
Above mentioned idea is just my personal opinion on how we can potentially solve this problem. Other solutions are highly encouraged and will be happily considered if they are successful in solving the above mentioned problem.