Instead of turning off all "expensive optimizations" in gcc,
I found I can achieve consistent calculation results on
64-bit ARM processors by turning off just the "fp contract"
optimization.
This optimizer is actually supposed to produce *more* accurate
results, but the effect is to produce results that differ
across languages/processors. For the sake of the unit tests,
I have decided it must be turned off in ctest and generate.
It turns out that in gcc 9.3.0 on aarch64, the "expensive optimizations"
were causing incorrect calculations. I resumed using -O3 but with
expensive optimizations turned off. Now all unit tests pass again.
Fixed some build warnings that occur on various gcc
optimization levels, and only on this version of gcc.
For now, build ctest with fewer optimizations: -O1
instead of -O3. This is because -O3 and -O2 cause
excessive errors in 'ctest diff' of the order
1.0e-9, where I usually get 1.0e-12. I will have to
come back and figure out exactly which optimization(s)
are causing the problem and turn them off specifically.
This also means I need to document the dangerous optimizations
for people who are using the C version of Astronomy Engine.
When 'ctest diff' fails because of excessive numeric error,
print out the two lines of input text that had the worst
numeric error. This really helps on the Raspberry Pi
where memory is at a premium, and it's hard to open the
full output files using vi.
