fix(cli): align row 2 imported workflow parity

[mstykow]

Summary

• fix imported replay directory retention so row-2 --from-json workflows keep matching standalone directories and synthetic replay roots
• remove null-language per-resource tallies from row-2 output and extend xtask compare-outputs to compare summary/tallies/facet workflow output meaningfully
• mark CLI workflow verification scorecard row 2 verified after rerunning representative imported-JSON compare targets against ScanCode

Scope and exclusions

• Included:
  • row-2 imported workflow parity for --classify, --summary, --tallies*, --facet, and --license-clarity-score
  • replay directory synthesis/retention fixes in scan_result_shaping
  • compare harness normalization for row-2 sections and resource-level tallies
• Explicit exclusions:
  • new row-3 --package-only verification work
  • broader tally-content normalization beyond the major row-2 workflow bugs fixed here

Intentional differences from Python

• Remaining row-2 differences are limited to normalization/detail choices in tally content (for example preserved year/context in copyright tallies) rather than missing workflow sections, dropped replay resources, or broken classify/facet output.

Follow-up work

• Created or intentionally deferred:
  • continue with scorecard row 3 (--package-only) in a follow-up PR