Regression harness fails due to non-initialized variables; timing mismatch; namelist parameters are missing

The current source code does not allow creating the test harness and a candidate baseline test result. The following
independent problems make the results scientifically unsuitable as references:

1. several output fields contain values from uninitialized Fortran arrays;
2. the fuel-strip and observed-perimeter cases never ignite because the first
   model advance starts at 4 s and misses their ignition windows;
3. observed-perimeter data are installed after propagation has already begun;
4. atmospheric fields written at an output time contain the preceding forcing
   record;
5. the OpenMP-4 test receives only one computational tile;
6. required namelist values are missing from either input or MPI broadcast.

The harness must also require evidence of fire propagation before candidate
validation can pass. Without that check, initializing the affected arrays
would remove the current MPI comparison failure while still allowing two inert
fire cases to become baselines.

The file describe the 8 issues:

CFBM-RegressionTests-PreBaseline-Issues-v1.md

CFBM-RegressionTests-PreBaseline-Technical-Findings-v1.md

Start by reading CFBM-RegressionTests-PreBaseline-Issues-v1.md and CFBM-RegressionTests-PreBaseline-Technical-Findings-v1.md, then trace the Fortran regression harness, candidate validation, and namelist/MPI setup described there. Done means all eight listed problems are resolved, fire propagation is required before validation, and a scientifically valid candidate baseline can be created.