t is immediate on 64bit boxes and so plays nicely with the GC write barrier.
Unfortunately, private Core_int63.t is necessary for the compiler to optimize
Returns Neg, Zero, or Pos in a way consistent with the above functions.
The limits of t are chosen to allow conversion to and from float spans with
microsecond precision. This property supports Core.Std.Timing_wheel_float in
particular. See also Core.Std.Time.
overflows silently on min_value
Fast, implemented as the identity function.
Will raise on 32-bit platforms. Consider to_int63_ns instead.