#4744 Maxima doesn't know that zeroa < any positive number

[David Scherfgen]

See: https://sourceforge.net/p/maxima/bugs/4688/

Adding the following assumptions seems like a way to go:

(learn `((mlessp) $zeroa ,+least-positive-flonum+) t)
(learn `((mgreaterp) $zerob ,+least-negative-flonum+) t)

But doing so results in a stack overflow in this integral:

assume(l > 0, x > 0)$
integrate(integrate((acos(x/l)+acos(y/l)-%pi/2)/(2*%pi),y,0,sqrt(l^2-x^2)),x,0,l);