mpich: update to support gcc11

[mascguy (Christopher Nielsen)]

When this was last tested earlier in 2021, compilation failed for various reason.

On ARM, the build dies immediately after the configure phase is completed, with errors like the following:

libtool: compile:  /opt/local/bin/gfortran-mp-11 -Isrc/binding/fortran/use_mpi -Isrc/binding/fortran/use_mpi_f08 -pipe -m64 -O2 -Jsrc/binding/fortran/use_mpi_f08 -c src/binding/fortran/use_mpi_f08/mpi_f08_types.f90  -fno-common -o src/binding/fortran/use_mpi_f08/.libs/mpi_f08_types.o
src/binding/fortran/use_mpi_f08/mpi_f08_types.f90:355:17:
  355 |     real(real128),dimension(..) :: x
      |                 1
Error: Kind -1 not supported for type REAL at (1)
src/binding/fortran/use_mpi_f08/mpi_f08_types.f90:385:20:
  385 |     complex(real128),dimension(..) :: x
      |                    1
Error: Kind -1 not supported for type COMPLEX at (1)

But for Intel, there are complaints about array element alignment:

/bufattachf.c  -fno-common -DPIC -o src/binding/fortran/mpif_h/.libs/lib_libpmpi_la-bufattachf.o
:info:build In file included from ./src/mpid/ch4/netmod/include/../ofi/ofi_impl.h:10,
:info:build                  from ./src/mpid/ch4/netmod/include/../ofi/ofi_am.h:8,
:info:build                  from ./src/mpid/ch4/netmod/include/../ofi/netmod_inline.h:9,
:info:build                  from ./src/mpid/ch4/netmod/include/netmod_impl.h:1644,
:info:build                  from ./src/mpid/ch4/include/mpidch4.h:322,
:info:build                  from ./src/mpid/ch4/include/mpidpost.h:10,
:info:build                  from ./src/include/mpiimpl.h:241,
:info:build                  from src/binding/fortran/mpif_h/mpi_fortimpl.h:117,
:info:build                  from src/binding/fortran/mpif_h/bufattachf.c:8:
:info:build ./src/mpid/ch4/netmod/include/../ofi/ofi_types.h:326:5: error: size of array element is not a multiple of its alignment
:info:build   326 |     MPIDI_OFI_cacheline_mutex_t mutexes[MAX_OFI_MUTEXES];
:info:build       |     ^~~~~~~~~~~~~~~~~~~~~~~~~~~

We should investigate further - including any potential upstream patches - to see if it's easily fixable.

[mascguy (Christopher Nielsen)]

Issues are now resolved, and mpich-gcc11 is officially in the wild. Closing.