[Performance] shootout.binarytrees when implemented with gc is 10x slower than C# on .NET 2.0

[Craig Black]

> C# on .NET 2.0     :  1.902 sec
> DMD 0.173 - malloc(original version)  :  5.189 sec
> C# on Mono 1.1.18 :  6.630 sec
> DMD 0.173 - full GC : 19.720 sec

I see that C# GC is faster than DMD's malloc.  How is this possible?  Did
Microsoft figure out how to make the GC really fast, or is D's malloc slow?
I wonder how D's malloc compares to nedmalloc.  It is a very fast
cross-platform allocator.  If it is the case that D's malloc is slow, why
don't we adopt a public domain allocator that is real fast?  It would be
very easy to do.

-Craig