Use macros for operating on int64_t quantities, even in platform code:
----------------------------------------------------------------
The macros allow int64_t quantities to be manipulated, even if they
are implemented as structures.

The win32 code, although platform specific, is used as a basis for
other platforms.
Macros should be used for 64-bit arithmetic throughout the code, even
for platform specific code. Thius makes porting much easier on platforms
without 64-bit integers.

The macros will reduce to normal 64-bit arithmetic on Win32, with
no loss in performance, and only a little loss in code readablity.

As a test, it should be possible to
compile Win32 code using a structure for int64_t quantities.

E.g.
typedef struct {
   __int64 val;
} int64_t;

#define ll_add(x, y) ((x).val + (y).val)

etc.

(jni_invoke_native may need changing for return types, but the rest of
the product should compile fine).