Hello

This patch adds support for the OpenMP 'declare variant' directive in 
Fortran (C/C++ support is already present). For most part, this is a 
straightforward port of the C frontend code. I have ported all the 
c-c++-common/gomp/declare-variant-*.c tests to Fortran in 
gfortran.dg/gomp/, skipping over any tests involving inline assembly. I 
have also ported libgomp.c/target-42.c to Fortran as 
libgomp.fortran/declare-variant-1.f90.

I have moved the c_omp_check_context_selector and 
c_omp_mark_declare_variant functions from c-family/c-omp.c into 
omp-general.c and removed the 'c_' prefix, as these functions can be 
reused for Fortran.

Since the Fortran FE parses code first before translating it into tree 
form, fatal parser errors in a source file will result in errors in the 
same source file that are detected at translation time not being 
displayed (as compilation does not reach that stage). I have therefore 
separated out the translation errors in declare-variant-2.f90 into 
declare-variant-2a.f90.

In secion 2.3.1 of the OpenMP 5.0 spec, it says:

3. For functions within a declare target block, the target trait is 
added to the beginning of the set...

But OpenMP in Fortran doesn't have the notion of a declare target 
_block_ (i.e. the #pragma omp declare target/#pragma omp end declare 
target form), only the !$omp declare target (extended-list)/[clause] 
form (which C/C++ also has). The C FE differentiates between the two (it 
applies an 'omp declare target block' attribute for the first, an 'omp 
declare target' for the second) but only the first matches against the 
'target' construct in a context selector. I opted to match against 'omp 
declare target' for Fortran only, otherwise this functionality won't get 
exercised in Fortran at all. This difference is tested in test3 of 
declare-variant-8.f90, which I have XFAILed for now.

The Fortran 'declare variant' directive can also optionally take the 
name of the base procedure - this is implemented by moving any 'declare 
variant' directives with a matching base name in the parent namespaces 
of a function into the function namespace itself, such that at 
translation time it appears as if the directive was placed in the base 
procedure. This is tested in declare-variant-15.f90.

Bootstrapped on x86_64, and the gfortran testsuite and libgomp.fortran 
tests run with no regressions. Okay for trunk?

Thanks

Kwok Yeung