[vm/ffi] Test FFI with high heap

[sjindel-google]

Currently native heap addresses are freely cast between void*, intptr_t and size_t. We need to make sure there are no surprises when the heap lives in the upper half of the address space.

[mkustermann]

There's also some interesting questions on 64-bit runtimes:

final p = Pointer<Int8>.fromAddress(0);
p[-1]; // Should be disallowed (or treat the index as uint64)?
p[0x8000000000000000]; // Same here, since index is -9223372036854775808

final p2 = Pointer<Int8>.fromAddress(0x8000000000000000); // Should be allowed.

[sjindel-google]

I don't see why p[-1] should be disallowed. I think the C interpretation should be (p.address + -1 * elementSize).load().

And the same for the rest. But, I intended this issue to be specifically about our handing of malloc returning large pointers. So it's obsolete after #38520