FIX: introduced a bug in a loop by uint8_t inside crapto1.c

CHG: textual helptext change in "hf mf nested"
CHG: loop inside "nested". Still not fast enough. I wonder if qsort can be exchanged for radixsort or bucket sort?

client/nonce2key/crapto1.c

@@ -188,11 +188,11 @@ struct Crypto1State* lfsr_recovery32(uint32_t ks2, uint32_t in)
 	recover(odd_head, odd_tail, oks, even_head, even_tail, eks, 11, statelist, in << 1, bucket);
 
 out:
+	for (uint32_t i = 0; i < 2; i++)
+		for (uint32_t j = 0; j <= 0xff; j++)
+			free(bucket[i][j].head);
 	free(odd_head);
 	free(even_head);
-	for (uint8_t i = 0; i < 2; i++)
-		for (uint8_t j = 0; j <= 0xff; j++)
-			free(bucket[i][j].head);
 	return statelist;
 }