Generalize skewed diagonal approach to rectangular matrices

[ashvardanian]

Currently, the anti-diagonal approach for Levenshtein distance is only implemented for square matrices.
It should be easy to generalize it to rectangular ones, opening opportunities for more aggressive vectorization.

StringZilla/include/stringzilla/stringzilla.h

Lines 2010 to 2013 in 0642318

	// TODO: Generalize to remove the following asserts!
	sz_assert(!bound && "For bounded search the method should only evaluate one band of the matrix.");
	sz_assert(shorter_length == longer_length && "The method hasn't been generalized to different length inputs yet.");
	sz_unused(longer_length && bound);