Lines with column names consisting of national letters
were wrongly formatted in "mysql --table" results:

mysql> SELECT 'xxx xxx xxx' as 'xxx xxx xxx';
+-------------------+
| xxx xxx xxx |
+-------------------+
| xxx xxx xxx       |
+-------------------+
1 row in set (0.00 sec)

It happened because in UTF-8 (and other multibyte charsets)
the number of display cells is not always equal to the number
of bytes of the string.

Data lines (unlike column name lines) were formatted correctly,
because data lines were displayed taking in account number of
display cells. This patch takes in account number of cells when
displaying column names, the same way like displaying data lines does.

Note: The patch is going to be applied to 4.1.
Test case will be added after merge to 5.0,
into "mysql.test", which appeared in 5.0.

mysql.cc:
  Adding column name allignment using numcells(),
  the same to data alignment, which was implemented earlier.