Removed unpredictable test from rpl000001.test

Fixed rpl000001 to handle the most common mysql-test-run ports.

Don't reconnect client on close

Fixed bug with float columns

    Added a stern warning in a visible (in my opinion) place to make sure
    to compile with -fno-exceptions when using gcc
    Wrote a new paragraphs on OS comparison/choice in the Which OS section

Bug fix for Dec OSF on Alpha.

Minor bug fix for server RPM.
Tweak RPM support for mysqld-max.

Portability fix for Innodb

log0log.c	Change Innobase text to InnoDB
srv0start.c	Change Innobase text to InnoDB
srv0srv.c	Change Innobase text to InnoDB
ha_innobase.cc	Change Innobase text to InnoDB

Deleted a lot of old benchmark runs

    do REAL_EMAIL magic
mysql-test/mysql-test-run.sh
    added show_failed_diffs
mysql-test/r/bench_count_distinct.result
    tired of waiting :-)
mysql-test/r/rpl000001.result
    new test results after implementing Monty's Hack (TM)
mysql-test/t/bench_count_distinct.test
    tired of waiting
mysql-test/t/rpl000001.test
    implemented Monty's Hack (TM) to kill a query in the middle of update
    without generating a table with 30,000 rows. This will also avoice the
    timing uncertainty

    reversed incorrect temporary change

start slave thread if master.info is present and server_id is set
updates for rpl000015 and rpl000016 to remove master.info to make sure
master info struct does not get auto-initialized