Replication: small bug fixes in mysql-test-run.sh and in replication code (more reinitializations)

+ fix for #70 
+ test result fix for LOAD DATA LOCAL bug #82


client/mysqltest.c:
  Removed wrong comment and useless sleep
mysql-test/mysql-test-run.sh:
  Fix $slave_datadir variable because this
    if [ -n "$1" ] ; 
     slave_datadir="var/$slave_ident-data/"
    else
      slave_datadir=$SLAVE_MYDDIR
  was not coherent and led to problems of master.info not being deleted.
  Delete relay-log.info too.
mysql-test/r/rpl000001.result:
  SUM is 1022 (1021 was when we had the LOAD DATA LOCAL bug #82).
  It's logical to have an even number as we load the same file twice.
sql/slave.cc:
  When we RESET SLAVE, clear rli->master_log_* to give a good display in SHOW SLAVE STATUS ;
  seeing the old values may confuse the user.
  When we START SLAVE, clear rli->master_log_name, not only rli->master_log_pos.
sql/sql_repl.cc:
  When we RESET SLAVE, clear mi->master_log_* to give a good display in SHOW SLAVE STATUS (bug #70)

mysql-test/r/rpl000001.result

@@ -32,7 +32,7 @@ n
 2
 select sum(length(word)) from t1;
 sum(length(word))
-1021
+1022
 drop table t1,t3;
 reset master;
 slave stop;