Removing NO_FLOAT from gnu make source

[Paul Smith]

Hi all; I believe that the NO_FLOAT compile option in GNU make is not
really relevant anymore and I'd like to remove it just to clean up a bit
of code.  I don't know of any system where you can't declare variables
of type double or float and I'm not aware of any situation in which this
would make a difference in the performance of make, even for systems
without a decent FPU: the only places it's used is during load
computation ("make -l") and a few places dumping rules ("make -p").

If anyone has any contrary opinions please let me know.