mpfr is multiple precision (arbitrary precision if you prefer). IMO it's too much for FloatMathPlugin. CRLIBM (correctly rounded mathematical libray) could be a good modern alternative to fdlibm (from INRIA too).
CRLIBM is also LGPL https://hal-ens-lyon.archives-ouvertes.fr/ensl-01529804/file/crlibm.pdf