I believe that's the wrong questions...

If you really want / need to compare speed, the only way to answer that is to program the two methods and to measure them in a reliable way (beware, in Java, of naive micro-benchmarks! You have to take in account garbage collection, Jit compiler kicking in, and so on).

From a more technical and theoretical point of view / analysis, you have to take a look at the source code of Processing, particularly PGraphics, PShape, PShapeSVG and other variations.
You can see, for example, that when PShape draws primitives, it uses vertex(), quadVertex, bezierVertex, etc.
So, fundamentally, they use the same rendering technique, but PShape has a bit more ceremony around it, perhaps using more memory, etc.

Now, about the wrong question: should I have to draw your image in Processing, I would, certainly, use laodShape(): it would take perhaps one hour to draw it in Inkscape, while it would take much more to code it. And it is a relatively simple shape, and I am used to compute coordinates in my mind, designing such shapes...
You can also hand-code Processing calls using coordinates from a SVG file, but it is kind of useless, and, again, the gain of time is probably very low.
So, the right question is: what is the more convenient to draw such shape? The answer is, obviously, loadShape().

Although doing it by hand once can be a good exercise, a bit like coding in assembly language instead of C: it gives a good understanding of the underlying mechanisms, and helps coding at a higher level.

In the past, when computers were very slow (CPU ran at 1MHz, yes, mega), it was more performant to use assembly language, and even to unroll loops, using copy / paste the number of lines on screen instead of using a loop (it has an overhead, incrementing a counter, comparing the value to the limit) to gain some microseconds.
Today, the gain would be in nanoseconds, so unless it is carefully benchmarked to prove there is a real gain, it is often not worth the effort.