*	pointer within the string (>= start).  When (src == start), it
 *	returns start. When (src > start), it returns a pointer (< src)
 *	and (>= src - TCL_UTF_MAX).  Subject to these constraints, the
 *	routine returns a pointer to the earliest byte in the string that
 *	starts a character when characters are read starting at start and
 *	that character might include the byte src[-1]. The routine will
 *	examine only those bytes in the range that might be returned.
 *	It will not examine the byte *src, and because of that cannot 
 *	determine for certain in all circumstances whether the character
 *	that begins with the returned pointer will or will not include
 *	the byte src[-1]. In the scenario, where src points to the end of
 *	a buffer being filled, the returned pointer point to either the
 *	final complete character in the string or to the earliest byte
 *	that might start an incomplete character waiting for more bytes to
 *	complete.
 *
 *	Because this routine always returns a value < src until the point
 *	it is forced to return start, it is useful as a backward iterator
 *	through a string that will always make progress and always be
 *	prevented from running past the beginning of the string.
 *
 *	In a string where all characters are complete and properly formed,
 *	and the value of src points to the first byte of a character, 
 *	repeated Tcl_UtfPrev calls will step to the starting bytes of
 *	characters, one character at a time. Within those limitations,
 *	Tcl_UtfPrev and Tcl_UtfNext are inverses. If either condition cannot
 *	be met, Tcl_UtfPrev and Tcl_UtfNext may not function as inverses and
 *	the caller will have to take greater care.
 *
 * Results: