Reduce numeric-identifier noise in threat actor similarity report

[adulau]

Motivation

• Threat-actor names frequently reuse a textual stem with different numeric identifiers (e.g., apt 28 vs apt 29), which creates noisy high-similarity matches; the change aims to reduce these false positives while keeping comparisons focused on non-numeric content.

Description

• Add strip_numeric_tokens() to remove digit sequences from normalized names and numeric_tokens() to extract numeric token sequences.
• Update find_similar_name_pairs() to skip candidate pairs when both names contain numeric tokens, share the same non-numeric stem, but have different numeric sequences.
• Document the numeric-variant filtering behavior in the function docstring and keep all other similarity logic (length pre-filter, SequenceMatcher) unchanged.
• No changes to the CLI or output formats were made; markdown and JSON outputs remain the same.

Testing

• Ran python3 -m py_compile tools/threat_actor_similarity_report.py which succeeded.
• Ran python3 tools/threat_actor_similarity_report.py --help | head -n 20 to validate CLI help which printed successfully.
• Executed python3 tools/threat_actor_similarity_report.py --max-results 5 --markdown-output /tmp/threat_report.md --json-output /tmp/threat_report.json which produced a report and JSON file, printing analysis summary (reports written) with no errors.

---

Codex Task