mempool: fix benchmark CheckTx for hitting the GetEvictableTxs call (#7796)

Based on the discussion in #7723, make the CheckTx benchmark exercise 
GetEvictableTxs which is one of the critical paths in CheckTx.

After profiling the test, the sorting will occupy 90% of the CPU time in CheckTx.
In the test it doesn't count the influence of the preCheck, postCheck, and 
CheckTxAsync when the mempool is full.

internal/mempool/mempool_bench_test.go

@@ -14,8 +14,13 @@ func BenchmarkTxMempool_CheckTx(b *testing.B) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
+	// setup the cache and the mempool number for hitting GetEvictableTxs during the
+	// benchmark. 5000 is the current default mempool size in the TM config.
 	txmp := setup(ctx, b, 10000)
+	txmp.config.Size = 5000
+
 	rng := rand.New(rand.NewSource(time.Now().UnixNano()))
+	const peerID = 1
 
 	b.ResetTimer()
 
@@ -26,9 +31,11 @@ func BenchmarkTxMempool_CheckTx(b *testing.B) {
 		require.NoError(b, err)
 
 		priority := int64(rng.Intn(9999-1000) + 1000)
-		tx := []byte(fmt.Sprintf("%X=%d", prefix, priority))
+		tx := []byte(fmt.Sprintf("sender-%d-%d=%X=%d", n, peerID, prefix, priority))
+		txInfo := TxInfo{SenderID: uint16(peerID)}
+
 		b.StartTimer()
 
-		require.NoError(b, txmp.CheckTx(ctx, tx, nil, TxInfo{}))
+		require.NoError(b, txmp.CheckTx(ctx, tx, nil, txInfo))
 	}
 }