Clean up make-float-aux due to #277

In make-float-aux, we have this bit of code:

	  (when (and (zerop result) (not (zerop number)))
	    ;; The number we've read is so small that it gets
	    ;; converted to 0.0, but is not actually zero.  In this
	    ;; case, we want to round such small numbers to
	    ;; least-positive-foo-float.  If it's still too small, we
	    ;; want to signal an error saying that we can't really
	    ;; convert it because the exponent is too small.
	    ;; See CLHS 2.3.1.1.
	    (let ((float-limit (ecase float-format
				 ((short-float single-float)
				  least-positive-single-float)
				 (double-float
				  least-positive-double-float)
				 #+double-double
				 (double-double-float
				  ext:least-positive-double-double-float))))
	      (if (>= (* 2 ratio) float-limit)
		  (setf result float-limit)
		  (error _"Underflow"))))

However, since #277 (closed) was fixed, we return a correctly rounded value for result. Hence the code testing ratio against float-limit isn't needed. If result is 0 and number is not zero, then the number definitely can't be represented as a float. We can remove all of that float-limit stuff and just throw an error.