Function data hash could just pack the inputs into a field

[sirasistant]

struct FunctionData {
    // First four bytes of the abi encoding 
    // of a function. 
    selector : FunctionSelector,
    is_internal : bool,
    is_private : bool,
    is_constructor : bool,
}

This fits into a single field. We could just not hash and implement a pack method, and use it instead.

[iAmMichaelConnor]

I guess it depends which is more efficient (across both the kernel and aztec.nr), constraint-wise: packing vs hashing.