About DecomposableTable trait API

[DoHoonKim8]

Currently there are two methods that maps memory_index to subtable_index and chunk_index.
This seems not straightforward, it would be better to return the indices of subtables that is used by memories for each chunk. But this requires that there are no memories in the same chunk that uses the same subtable. If this conjecture turns out to be correct, then refine the API as mentioned.

pub trait DecomposableTable<F: PrimeField>: Debug + Sync + DecomposableTableClone<F> {

    ...

    fn memory_to_subtable_index(&self, memory_index: usize) -> usize;

    fn memory_to_chunk_index(&self, memory_index: usize) -> usize;
}

[DoHoonKim8]

Memo: remove DecompsableTable trait, implement memory checking algorithms(ex. offline memory checking for Lasso, LogUp+GKR, halo2 subset argument) to work behind original meta.lookup API, and compose it with halo2 gate constraints. This approach seems better for Jolt implementation, since we split the indices and play with them around, this requires various constraints. It's better to have witness columns for subtable lookups and constraint them manually.