ip_4 must handle double-precision arguments

[edwardhartnett]

We should not need ip_d builds. Double-precisions libraries should not be required.

@webisu @AlexanderRichert-NOAA @AlysonStahl-NOAA

[AlexanderRichert-NOAA]

Do you envision the _4 build using double precision under the hood for doing the interpolation, or would it all be 4 byte?

[edwardhartnett]

Double precision should be used when double arguments are used.

In other words, it should accept arguments as 4-byte reals, and as 8-byte reals, and should use 4-byte or 8-byte math accordingly.

So one library, the _4, should be able to handle all _4 and _d needs. We would then just build the _4 and everyone would use that, there would be no _d build.

What do you think?