Fix CrayPat profiling for `lfric_atm`

### Are there any linked Issues or Pull Requests?

#103

### Brief description

Currently, CrayPat/perftools profiling only works with `gungho_model` on a single node. We want to be able to use these profiling tools for all applications and include perftools in a profiling suite.

### Further details of the issue.

When trying to profile `lfric_atm` with perftools, the error returned states that XIOS cannot open a NetCDF file in parallel. I have been able to get around this issue by instrumenting XIOS and `lfric_atm` with pre-instrumented versions of the `pnetcdf` library and a group of libraries contained under the perftools trace group named `io`:

`pat_build -u -g pnetcdf -g io xios_server.exe`

This allowed the `lfric_atm` and `xios_server` executables to be profiled.

> [!NOTE]
> 1. This was a traced run where all subroutine calls were logged. This is instead of sampling where perftools records the most called routines and then creates a newly instrumented executable that only traces the most called routines.
> 2. A traced run tracking every routine takes much longer to complete.
> 3. This was not able to profile MPI calls and no `MPICH_RANK_REORDER` file was generated
A possible solution could be to add a version of XIOS instrumented with the trace groups to the LFRic stack and update the `lfric_atm` perftools command to match.

No source file or test is named. Start by reviewing the perftools command for `lfric_atm`, the working `gungho_model` setup, and the linked issue #103; use the XIOS and `lfric_atm` profiling commands described here as the entry point. Done means profiling works for `lfric_atm` and other applications without the parallel NetCDF error and can be included in the profiling suite.

Written by the indexing model from the issue text.