Not equal using `equal_cells`

[lognorman20]

This PR aims to optimize #119. Implements equal_cells in is_zero in an effort to reduce constraints. The result of doing this are presented in this table:

Backend	Constrains w/o not_equal	Constraints w/ not_equal
R1CS	10	12
Kimchi	19	24

Although the number of constraints went in the wrong direction, I have an idea for improvement. First is implementing more bitwise operations to avoid adding constraints. Functions such as equal, is_zero_cell, and even arithmetic operations like sub and mul can be implemented using bitwise operations that avoid creating unnecessary internal variables. As the language continues to expand, more operations will come such as bitshifts and xor which will enhance the possibilities to optimize. Moving the workload to bitwise operators to reduce constraints might just be moving the problem to another area though and I'm curious to hear what other ideas there are.

[lognorman20]

Another aspect I noticed is that the # of constraints for equal and not equal are roughly the same. Definitely some improvement can be done for both

[mimoo]

thanks for the experiment here! I can close this one as we merged the other!