Removed unneded macro

[saghul]

No description provided.

[bnoordhuis]

Oh, by the way: *unneeded

[chqrlie]

We originally used JS_FLOAT64_NAN instead of NAN to ensure the NaN values were normaiized even on architectures were NAN does not always expand to a normalized value. This may be necessary for 32-bit NAN boxing on some exotic targets. Defining JS_FLOAT64_NAN appropriately for these targets would solve the problem.

Removing such wrappers increases the adherence on target specific behavior. Trying to keep these to a minimum helps with portability.

[bnoordhuis]

I wondered about that. I suspected it was to signal that there's a "canonical" NaN but then we don't normalize NaNs (or not consistently anyway), rendering it moot. It mostly seems like baggage at this point.

[saghul]

@chqrlie Do you know what those excotic targets are?