Caching field normalize and bigfield self_reduce

[Rumata888]

Currently we just add constraints and return new elements when we normalize field elements and reduce bigfield elements. This means that a call to the same field element twice to normalize it can produce 2 sets of identical constraints. Ideally, we would cache the result and just return it

[elijahhampton]

Is this issue open to work? @Rumata888

I hesitated due to it being opened in the mirrored repository vs aztec-packages