Rewrote ODBC's read-data-in-chunks to handle multibyte characters.

Rewrote ODBC's read-data-in-chunks to handle multibyte characters.

This function is called on large (greater than +max-precision+) fields
and was previously converting to lisp characters one at time with no
regard for encoding.  The new method uses 2x more memory as it uses an
extra copy of the string (in pieces) but at least it returns correct
data.

Added a test case that shows the problem (now fixed).