If you have a giant tear-drop shape you are still going to need enough fin area in the back to move the CP aft, or enough ballast in the nose to move the CG forward, or both. If your design has unconventional shapes, or even normal shapes but oriented in unusual ways, you are not going to find a cook-book method (like Barrowman) that applies. If you have portions of the design (especially fins) immersed in turbulent flow that is produced by fore-bodies (or other fins) that is also going to complicate things.
Estes produced several kits with Star-Wars themes that were made stable by attaching the desired object/shape to a long body tube. This serves as a boom or arm to help hold nose ballast at a greater distance out the front of the rocket, making it possible to use less ballast weight. You could possibly do something similar, whether using a body tube or using a rod (1/4 inch wooden dowel rod?) to hold a small nose pod with ballast, hidden a bit (or at least, made less visually insulting) by painting it black.
A safe bet is always to use tons of nose ballast but that has obvious performance penalties. Some people use oversize tail fins and (sort of) hide them by using clear plastic sheet. Other people use a base plate as a drag-producing device (which can also be hidden slightly by making them from clear plastic sheet.
Before you invest a lot of time and work in a shape that cannot be made stable, you might do a boilerplate version with a rough representation of the size, shape, and weight. See if that version flies OK before you put a lot of energy into building a deluxe version.
I think you will find that CFD tools will not work reliably on unconventional shapes either (at least, for most of the simplified CFD methods available for use with rocketry). Part of the premise for being able to use computer procedures to model airflow is that the airflow is assumed to be generally smooth and well behaved and that the surface of your rocket is smooth and has gentle changes in contour. It is fairly easy to overwhelm the CFD internal mathematics if you are trying to analyze too many features, or odd features, or zones of stalled/separated/severely turbulent airflow near your rocket. Confidence in analysis results degrades accordingly.
You could build a scale model and test it in a wind tunnel. That is going to be way more trouble than its worth, plus afterward you will still be faced with the problem of scaling the results back up to full scale (unless you test at full size and full q).
Short answer: there isnt an easy way to determine this. Trial and error works as well as anything else. A boilerplate test vehicle, launched away from crowds, is likely to be your easiest approach.