Sort outputs

Based on conversation in #41, outputs in `--archive-dive` and `--no-archive-dive` modes should come out in a consistent and predictable order.

Current behavior is to follow `os.walk` with files first then subfolders and not alphabetically. With archive-dive ON, archives are written last so an archive ends up separated from its sibling files.

Conceivably, a user would want outputs sorted on the hash to see identical files grouped together regardless of their location in a collection, but that type of sorting seems more appropriate for downstream use.

This could have implications for parallel hashing (#3), so this should probably be done in coordination with a PR that implements that. The upfront size reporting (part of #33) could also feed the parallel worker task planning so each worker would receive bins of equal 'weight' to hash through.

Start at the CLI entry point that handles --archive-dive and --no-archive-dive, then trace how os.walk results are emitted and how archives are deferred. Define and implement one consistent, predictable ordering for both modes, while checking the coordination concerns in issues #3 and #33 before considering the work done.

Written by the indexing model from the issue text.