Creating a new HDF5 nuclear data library for UQ

[Grego01-biot]

This PR adds generate_endf_covariance.py, a script that extends the existing generate_endf.py with optional MF=33 multigroup cross-section covariance processing. It produces the same HDF5 library as generate_endf.py but with covariance data (lower-triangular L factors) attached to each neutron file. It is important to note that this PR depends on openmc-dev/openmc#3911!

For each neutron ENDF evaluation, the script:

1. Runs the standard NJOY processing chain to produce cross-section data (identical to generate_endf.py)

2. Runs NJOY/ERRORR to produce multigroup relative covariance matrices on a 1500-group uniform-lethargy grid (1e-5 eV to 20 MeV). The default settings are covariance temperature at 293.6K, eigenvalue tolerance at 1e-10, temperatures for neutron cross-sections 250, 293.6, 600, 900, 1200 and 2500K.

3. Factors each covariance matrix as lower triangular and stores L in the HDF5 file under /<nuclide>/covariance/mf33/.

If ENDF evaluation lacks MF=33 for a nuclide, the HDF5 file is still written with standard cross sections. A warning is logged and processing continues.

The full library with covariance can be run using the following script:

python generate_endf_covariance.py --njoy /path/to/njoy -r viii.0 -d endfb-viii.0-hdf5

The impact on storage and time to generate this new library is in process and results will be added to this PR.