Add more context on too many signatures failure for transactions (tx_bad_auth_extra)

[theaeolianmachine]

We've gotten a few questions on why too many signatures (weight > threshold) will fail a transaction, mentioned in the Multi-Sig guide.

Nico gave context on this here:
stellar/stellar-protocol#120

Reason we did this was because of the quadratic nature of signature verification (which is the most CPU intensive thing we deal with before accepting transactions in consensus - the hints help but they don't help making the worst case). I agree this may be more burden than what it's worth.
It may make sense to reconsider this the next time we look into changing the fee structure: I think that there should still be an incentive to submit minimal transactions to the network.

We should add a note explaining that this exists in order to not unnecessarily increase the signature verification time before accepting transactions in consensus.