First, try googling "bicycle castor". You will find that the unfortunate clause in the Wikipedia article you cite is the only reference on the internet that equates caster with trail. The word "caster" when used with regard to vehicle suspensions refers to an angle, not a linear measurement. The confusion probably arose due to the fact that castor angle can be inferred from rake, trail, and wheel size, which are all linear measurements.
Trail is a guideline used by bicycle designers. It is neither necessary nor sufficient for stability: The authors of the Science article (
sources and preprint here) built a stable negative-trail "bicycle"; and they show that the eigenvalues for conventional bicycles are unstable above a critical speed. However, the suggestion that trail - the largest force affecting steering - is irrelevant to the stability of conventional bicycles is ludicrous. You can do the math and calculate eigenvalues as you alter trail, or simply ride a bike equipped with an adjustable trail for to see this for yourself.
To quote Jim Papadopoulos, "perhaps the main message is that our reasoning about how trail affects bicycle stability has been quite wrong for 120 years". This is not a denial that trail affects stability, but rather a search for a better model. Unfortunately, that better model has not yet yielded any insights of use to the designers of actual ridden bicycles.