With this commit, and the 2 previous ones, given mytest.cpp
```

int main()
{
    PJ* pj = proj_create(nullptr, "+proj=vgridshift +grids=us_nga_egm96_15.tif");
    for( int i = 0; i < 5*1000*1000; i++)
    {
        PJ_COORD coord;
        coord.lpz.lam = 0;
        coord.lpz.phi = 0;
        coord.lpz.z = 0;
        proj_trans(pj, PJ_FWD, coord);
    }
    return 0;
}
```

we get a x2 speedup

Before:
```
$ PROJ_LIB=data:$HOME/proj/PROJ-data/us_nga LD_LIBRARY_PATH=src/.libs  hyperfine --warmup 1 'taskset -c 11 ./mytest'
Benchmark #1: taskset -c 11 ./mytest
  Time (mean ± σ):      1.950 s ±  0.014 s    [User: 1.945 s, System: 0.005 s]
  Range (min … max):    1.937 s …  1.971 s
```

After:
```
$ PROJ_LIB=data:$HOME/proj/PROJ-data/us_nga LD_LIBRARY_PATH=src/.libs  hyperfine --warmup 1 'taskset -c 11 ./mytest'
Benchmark #1: taskset -c 11 ./mytest
  Time (mean ± σ):     984.4 ms ±   3.1 ms    [User: 977.0 ms, System: 7.2 ms]
  Range (min … max):   979.3 ms … 990.5 ms
```

Co-authored-by: Mike Taves <mwtoews@gmail.com>

The big size of coordinateoperation.cpp could require significant amount
of RAM to build it with -O2 level, and cause compiler crashes in some
environments.