This was originally done for testing purposes, but after #126315 we now
do testing through GitHub Actions and should be instead using the
optimization setting chosen by the user.
This PR aims to add the groundwork to test the precision of libc complex
functions against MPC. I took `cargf` as a test to verify that the infra
works fine.
- migrate more `-O3` to `${libc_opt_high_flag}`
- workaround a issue with `LLP64` in test. The overflow testing is
guarded by a constexpr but the literal overflow itself will still
trigger warnings.
Notice that for math smoke test, for some reasons, the
`${libc_opt_high_flag}` will be passed into `lld-link` which confuses
the linker so there are still some warnings leftover there. I can
investigate more when I have time.
I have commented out the test for `neg_zero`(creal) because :
1. real(neg_zero + 0.0i) equals zero.
2. real(neg_zero - 0.0i) equals neg_zero.
I am not sure if this is the intended behaviour.
[EDIT]
I have updated tests for `neg_zero` (creal) to be :
```
EXPECT_FP_EQ(func(CFPT(neg_zero - zero * 1.0i)), neg_zero);
EXPECT_FP_EQ(func(CFPT(neg_zero + zero * 1.0i)), zero);
```
because all three [gcc, clang and GNU MPC] also give the same result.
https://godbolt.org/z/hxhcn6aof
and it seems that it is indeed the correct behaviour since Imaginary
types are not supported yet, refer #113671
