net_processing: drop m_recently_announced_invs bloom filter

Rather than using a bloom filter to track announced invs, simply allow
a peer to request any tx that entered the mempool prior to the last INV
message we sent them. This also obsoletes the UNCONDITIONAL_RELAY_DELAY.

src/txmempool.h

@@ -708,6 +708,10 @@ public:
         return mapTx.project<0>(mapTx.get<index_by_wtxid>().find(wtxid));
     }
     TxMempoolInfo info(const GenTxid& gtxid) const;
+
+    /** Returns info for a transaction if its entry_sequence < last_sequence */
+    TxMempoolInfo info_for_relay(const GenTxid& gtxid, uint64_t last_sequence) const;
+
     std::vector<TxMempoolInfo> infoAll() const;
 
     size_t DynamicMemoryUsage() const;