UTXO selection algorithms select redundant UTXOs

[ImplOfAnImpl]

This is especially noticeable/reproducible in the config-broadcast no mode (introduced in #1904), in which the new transaction's summary is printed to the console without it being broadcast to the network.

E.g. when trying to create a new transaction I can see that my biggest UTXO (100k TML) is selected, but it's also accompanied by smaller UTXOs, which should not be needed.

There seem to be multiple reasons for this:

1. The selection algirithm to blame seems to be srd; in select_coins_srd we randomly select UTXOs, populating a heap, and in the end do:

   if selected_eff_value >= target_value {
       // Result found, add it.
       while let Some(group) = heap.pop() {
           result.add_input(&group.0, pay_fees)?;
       }
       return Ok(result);
   }
   

   So the reason for the smaller UTXO being initially selected is that the large UTXO was selected later. But then the above-mentioned code shouldn't put the entire heap into the result; only the top items necessary to meet the target should be put there.

2. In select_random_coins, several algoriths (like select_coins_srd) are executed and the best result is taken. And the best result selection is currently implemented as

   results.into_iter().min_by_key(|res| res.waste)
   

   In my case, the waste value was the same for all algorithms; and since select_coins_srd is executed first, it was always winning.
   So:

   • probably the best result should be selected first by waste, but if it's equal then the UTXO count should be taken into account as well.
   • perhaps waste calculation can be improved.