replace the sinsemillar note commitment with the poseidon commitment

[XuyangSong]

close #199 and resolve #123 incidentally

• Use Poseidon hash to commit the note, note_cm = poseidon_hash(note, rcm).
• As a result, the note commitment will be a field element instead of a point. Adjust the nullifier derivation from nf =Extract_P([PRF_{nk}(\rho) + \psi \ mod \ q] * K + note_cm) to nf = poseidon_hash((nk || \rho || \psi || note_cm)
• Remove sinsemilla commitment and the note bits decomposition
• Remove the confusing address(address = hash(app_data_dynamic || app_data_dynamic)) in the note. Both app_data_dynamic and nk_com directly go to the note commitment.

Potential improvement: There are seven field elements in Poseidon note commitment. If we composite the is_merkle_checked(bool) and value(u64 but may be extended to u128 later) to one element(is_merkle_checked * (1<<128) + value) in the circuit, we can save one round of Poseidon absorb. Is it worth it?

[bazzilic]

LGTM

[bazzilic]

we can save one round of Poseidon absorb. Is it worth it?

I think we can take note of this possibility and leave this optimization for the future.