szip formula should be removed

[Bo98]

The szip has a license that appears to not be fully open source: https://support.hdfgroup.org/doc_resource/SZIP/Commercial_szip.html.

There is now an open-source replacement however: libaec. We already ship it so we should migrate away form szip as soon as possible.

---

Progress:

• The two formulae have conflicts_with so we should probably think about handling that before any migration or users will get errors on upgrading. - libaec szip: change conflicts_with to keg_only #96933
• Migrating dependents:
  • hdf5@1.10 - hdf5@1.10 1.10.8 #96928
  • hdf5@1.8 - hdf5@1.8: depend on libaec instead of szip #96942
  • hdf5 / hdf5-mpi - hdf5 and dependents: downgrade and use libaec #97666
    • alembic
    • alpscore
    • armadillo
    • caffe
    • cdo
    • cgns
    • vtk
    • vtk@8.2
• Remove szip - szip: remove #98261

[cho-m]

I did see this when looking into hdf5.

I had opened #96928 relating to migrating hdf5@1.10 from szip to libaec, but can wait for fixing conflicts_with first.

I just opened #96933 to switch conflicts_with to keg_only.

---

Most other usage of szip is due to hdf5's dependency tree, but may need to also consider downgrading 1.13.0 back to 1.12.1.

[cho-m]

At this point, only need to migrate hdf formula and dependents in single PR before disabling szip.

However, we probably should decide if hdf should be downgraded to 1.12.1 beforehand or same PR, and the best procedure to do this. There are a number of dependents revision bumped so manually reverting files may break. Not sure if best option is to downgrade via version_scheme.