Ruling out arbitrary cost copy construction?

[Walter Bright]

Don wrote:
> The most important issues would be fixed with a small-value optimisation 
> (values x where -long.max <= x <= long.max stored in the struct).

Drat, I thought it already did that.

I think it's also important to make BigInt.sizeof == ulong.sizeof. To that end, 
you can use the least significant bit to mean "this is not a pointer", as in a 
63 bit integer << 1.

(Pointers will always be aligned.)