Increase the fp precision in dumped spec.

By default, C++ will output 6 fractional digits of floating point
values. This is insufficient for some of the RGG tests that requires
precision < 1e-6.

This CL uses hex representation for fp values in the generated spec
file, while human-readable values are also dumped as a comment.

Fixes: 144788409
Test: NNT_static_fuzzing and inspected the dumped spec
Change-Id: I43ed11e9f73cb387ed8894713fe79361322315ba
2 files changed