Failed tudat tests due to arm64 architecture

[FilippoOggionni]

When running the tudat tests through ctest, 5 of them fail (apparently due to differences in floating-point precision for different architectures).
I am using a MacBook Pro with M1 Apple Silicon processor. Failing test are:

The following tests FAILED:
         12 - test_basic_astro_OrbitalElementConversions (Failed)
         22 - test_basic_astro_TimeConversions (Failed)
         32 - test_basic_astro_EmpiricalAcceleration (Failed)
        143 - test_propagators_VariationalEquations (Failed)
        161 - test_basics_TimeTypes (Failed)

Copying and pasting from internal discussion (@DominicDirkx):

Test back and forth TCG<->TT at t=0, and expected value of TCG at TT=0..
    long double testTime = 0.0L;
    testTcg = convertTtToTcg< long double >( testTime);
    testTt = convertTcgToTt< long double >( testTcg );
    long double expectedTcg = -secondsSinceJ2000Synchronization * LG_TIME_RATE_TERM_LONG /
            ( 1.0L - LG_TIME_RATE_TERM_LONG);
    BOOST_CHECK_CLOSE_FRACTION( testTcg, expectedTcg, 2.0 * std::numeric_limits< long double >::epsilon( ) );
    BOOST_CHECK_CLOSE_FRACTION( testTt, testTime, 2.0 * std::numeric_limits< long double >::epsilon( ) ); 

are using the wrong test, I think (should be CHECK_SMALL, not CHECK_CLOSE_FRACTION).
Looking at this, I think that the long double is implemented as 64 bit, not 80 bits. Not a big problem, but good to be aware of

Full test output is attached here.