That idea of using a single non-vanishing component of a vector field as a representation of the scalar field, by f(x,y,z) --> f(x,y,z)*e[x], say, is nice, as the length of the arrow represents the magnitude of the scalar field and the direction represents its sign. So, in principle, there would be no additional information in need of using color.
Now yes, the help page ?plots,fieldplot3d is quite obscure in regards to coloring of the arrows. As far as I could see, the default coloring scheme (used when no color option is given) is not produced in the library code but later, in the renderer itself. Now, the procedure that controls/processes the admissible color option syntax is `plot/control`, and apparently an Array of color values could work, but it would need some experimentation.
So, as a shortcut or hack, I have bypassed it, again manipulating directly the plot structure. For instance, this command produces arrows all red:
> plots:-fieldplot3d([(x*y*z),0,0], x=-1..1,y=-1..1,z=-1..1,color=red):
and the information of color is contained in this COLOR function call:
> p:=plots:-fieldplot3d([(x*y*z),0,0], x=-1..1,y=-1..1,z=-1..1,color=red):
COLOR(RGB, 1.00000000, 0., 0.)
So, colours can be given to the arrows by substituting this COLOR function call with a HUE function call containing a sequence of float values for each arrow list, as in this test:
> subs(COLOR(RGB, 1.00000000, 0., 0.)=COLOR(HUE, colorseq),p);
PD. I have got this plot in Maple 18.02. It seems like a there is change/regression in Maple 2015 spoiling coloring.