Divide & Conquer divides, but doesn't conquer

[Adam D. Ruppe]

On Monday, 25 May 2020 at 13:00:59 UTC, Adam D. Ruppe wrote:
> Solid result though: 2s, 810MB.

That was using the new dmd master, more specifically it was 1.6s 
that I rounded up. With the released dmd, 1.9s, 750MB.

either way, the mixin helps here - but that may not be true if 
there was a variety of lengths. This test did fixed length 
because I wanted to specifically prove that arg length isn't as 
important as number of instantiations here...

we should prolly test a variety of lengths too, it could change a 
lot.