net_processing: Boost inv trickle rate

If transactions are being added to the mempool at a rate faster than 7tx/s
(INVENTORY_BROADCAST_PER_SECOND) then peers' inventory_to_send queue can
become relatively large. If this happens, increase the number of txids
we include in an INV message (normally capped at 35) by 5 for each 1000
txids in the queue.

This will tend to clear a temporary excess out reasonably quickly; an
excess of 4000 invs to send will be cleared down to 1000 in about 30
minutes, while an excess of 20000 invs would be cleared down to 1000 in
about 60 minutes.

Github-Pull: #27610
Rebased-From: 5b3406094f

src/net_processing.cpp

@@ -5594,7 +5594,9 @@ bool PeerManagerImpl::SendMessages(CNode* pto)
                     // especially since we have many peers and some will draw much shorter delays.
                     unsigned int nRelayedTransactions = 0;
                     LOCK(tx_relay->m_bloom_filter_mutex);
-                    while (!vInvTx.empty() && nRelayedTransactions < INVENTORY_BROADCAST_MAX) {
+                    size_t broadcast_max{INVENTORY_BROADCAST_MAX + (tx_relay->m_tx_inventory_to_send.size()/1000)*5};
+                    broadcast_max = std::min<size_t>(1000, broadcast_max);
+                    while (!vInvTx.empty() && nRelayedTransactions < broadcast_max) {
                         // Fetch the top element from the heap
                         std::pop_heap(vInvTx.begin(), vInvTx.end(), compareInvMempoolOrder);
                         std::set<uint256>::iterator it = vInvTx.back();