fix: remove system OpenMPI from NVHPC containers; enable IPO for all versions

- Remove openmpi-bin/libopenmpi-dev from NVHPC container apt install.
  The system OpenMPI (gfortran-compiled) was overriding the NVHPC-bundled
  MPI at runtime via ldconfig, causing ABI mismatch segfaults in all CPU
  tests (rayleigh_taylor_muscl crash in s_interface_compression).
- Remove NVHPC version exclusion range for two-pass IPO. Binary search
  identified s_compute_dt as the single function causing ICE on 24.11;
  add noinline directive to prevent its extraction into the inline library.

Author: sbryngelson

.github/workflows/test.yml

@@ -301,8 +301,7 @@ jobs:
         run: |
           apt-get update -y
           apt-get install -y cmake python3 python3-venv python3-pip \
-                             libfftw3-dev libhdf5-dev hdf5-tools \
-                             openmpi-bin libopenmpi-dev git
+                             libfftw3-dev libhdf5-dev hdf5-tools git
           git config --global --add safe.directory /__w/MFC/MFC
 
       # --- Standard build + test ---

CMakeLists.txt

@@ -262,11 +262,6 @@ if (CMAKE_BUILD_TYPE STREQUAL "Release")
     elseif (CMAKE_Fortran_COMPILER_ID STREQUAL "NVHPC")
         if (MFC_Unified)
             message(STATUS "LTO/IPO is not available with NVHPC using Unified Memory")
-        elseif (CMAKE_Fortran_COMPILER_VERSION VERSION_GREATER "24.11" AND CMAKE_Fortran_COMPILER_VERSION VERSION_LESS "25.9")
-            message(STATUS "LTO/IPO is not supported in NVHPC Version 24.11 to 25.9. Use >=25.9 or (<=24.11 && > 23.11) Performance will be degraded.")
-            set(NVHPC_USE_TWO_PASS_IPO FALSE)
-        elseif(CMAKE_Fortran_COMPILER_VERSION VERSION_LESS "23.11")
-            message(STATUS "LTO/IPO is not supported in NVHPC Version < 23.11. Use a newer version of NVHPC for best performance.")
         else()
             message(STATUS "Performing IPO using -Mextract followed by -Minline")
             set(NVHPC_USE_TWO_PASS_IPO TRUE)

src/simulation/m_time_steppers.fpp

@@ -625,6 +625,9 @@ contains
 
     !> Compute the global time step size from CFL stability constraints across all cells
     impure subroutine s_compute_dt()
+#ifdef __NVCOMPILER
+        !dir$ noinline
+#endif
 
         real(wp) :: rho  !< Cell-avg. density