New structs

[JMRyan]

Consider this uninspiring pair structs:

struct S1 {int x};
struct S2 
{
    int x
    this(int i) {x = i}
};

Note that no default constructor is allowed so that S2.init can have a
consistent value computed at compile time.

Now:

S1 a = S1();       // Quintessinal case works fine
S2 b = S2();       // Also works, D initializes b with S2.init
S2* c = new S1();  // Works, D initializes c* with S2.init
S2* d = new S2();  // Doesn't work: no default consructor

Instead of the last, we need:

S2* d = cast(S2*) GC.malloc(S2.sizeof);
d = S2();  // or: d = S2.init;

Is there any good reason why "S2* d = new S2();"
shouldn't be allowed? If allowed, D could initialize d* with S2.init.

S2 really isn't needed since S1(3) and S2(3) have the same effect.
Also, a final class would at least usually be just as good as a struct 
here.  But still, disallowing "S2* d = new S2();" seems decidedly
unnecessary, especially since "S2 b = S2();" already works.