fix build failure with nagfor7.0 because of using "complex" function

[sakamoti]

"complex" function which is not a fortran standard causes an error like below in nagfor.

(for example)
"Error: test_linalg.f90, line 445: Implicit type for COMPLEX in TEST_TRACE_CQP"

So, "complex" is replaced with "cmplx" which is one of the fortran standard function.
At the same time, I specified the "kind" value so that it has the same precision as the "cmplx" function argument.

[milancurcic]

Thank you @sakamoti! This is in part covered by #205, and to your PR also adds the explicit kind parameter to cmplx.

I suggest we wait to merge #205, then rebase this with master and merge. @certik does this sound reasonable?

[certik]

@sakamoti thank you for the fixes! I apologize that it took us longer to merge #205. We just did. Is there anything still failing in the latest master? If so, would you mind sending a PR or rebasing this one?

I apologize for this. Thank you again for our PR. Let me know if you need any help.

[sakamoti]

The latest version "48b28e8" resolved compilation error. But some "Questionable" message appeare from nagfor.

(example)
"Questionable: fortran-lang_stdlib/src/tests/io/test_savetxt.f90, line 106: Intrinsic function CMPLX with double precision argument and no KIND= argument returns single precision result"

So, adds the explicit kind parameter to cmplx is still meaningfull I think. How do you think about this?

If it is preffer to add kind parameter, I'll try it according to compiler message and send new PR.

[certik]

Yes, I think if the standard says that the result can be single precision (I think that's correct), we have to add the kind parameters. If you wouldn't mind doing it, we would really appreciate it. Thank you!

[sakamoti]

Please wait a moment, I just editing. (It is my first time to rebase...)
when finish, I will add the comment.

[sakamoti]

Rebase is over. Please review.

[certik]

I think the changes look great, thank you @sakamoti for fixing it!