fd_runtime_txn_lamports_per_signature does not match bank.rs:get_fee_for_message

[jumpsiegel]

bank.rs:get_fee_for_message looks like it gives preference to the lamports_per_signature in the blockhash queue over the lamports_per_signature contained in the nonce account. It looks like fd_runtime_txn_lamports_per_signature does exactly the reverse...

I noticed this while deep diving into the #1745 PR...

Right now, it is super difficult to change the lamports_per_signature but if we wrote a specific nonce tests(or fuzzer) and changed the stored lamports_per_signature in the advanced nonce accounts, it should generate different results between agave and fd

It would also just be interesting to randomize the lamports_per_signature associated with the blockhash in the blockhash queue for each txn generated. I bet this would find some interesting exceptions