fix the aggeration logic for multi-message with aux key

[drskalman]

From @AlistairStewart
One set of signers but many messages is:

$e(asig + t apk_1,g_2)=e(\sum_i H(m_i) +t g_1, apk_2)$

If each message m_1 has it's own set of aggregate public keys $apk_i,1$, $apk_i,2$ and linear combination factors t_i, you end up with:

$e(asig + \sum_i t_i apk_i,1 , g_2) = \sum_i e (H(m_i) + t_i g_1,apk_i,2)$

It's less useful to aggregate across messages with different signers in the g_1,g_2 scheme, because we have a big $apk_i,2$ for each message, so having one asig is not so useful.