moving clists out of kdata

[utzoo!decvax!harpo!npois!ucbvax!michael]

Looking at the dh code it seems that if the clists are not all in the
same unibus region (have the same extension bits) that the code will
not work properly.  The dhcsr dosn't track the extension bits
and therefore arbitrary or'ing of IENABLE may change them to the wrong
value.  I think the problems may actually not come up, but has someone
thought this through?  I am sure there would be problems
if a single clist element could span a region.  In dhxint the
&= ~XINT could reset the bits to 0 when the line has already incremented
them to 1.