Enhance advisory grouping 2172

[shivamshrma09]

Summary

Fixes #2172

This PR enhances the advisory grouping mechanism by adding two new
heuristics to the existing ComputeToDo pipeline in
compute_advisory_todo.py.

Changes

New issue types in ISSUE_TYPE_CHOICES (models.py)

• POTENTIALLY_RELATED_BY_ALIASES — for advisories from different
  datasources that share the same alias
• SIMILAR_SUMMARIES — for advisories with near-identical summaries

New pipeline steps (compute_advisory_todo.py)

relate_advisories_by_aliases()

• Iterates over all AdvisoryAlias objects
• If 2+ advisories from different datasources share the same alias,
  creates a POTENTIALLY_RELATED_BY_ALIASES todo
• Stores the shared alias in issue_detail
• Same-datasource advisories are skipped (expected, not interesting)

detect_similar_summaries()

• For each alias group with 2+ different datasources
• Compares all cross-datasource advisory pairs using
  difflib.SequenceMatcher
• If similarity ratio >= 0.8, creates a SIMILAR_SUMMARIES todo
• Stores similarity_score, datasource_a, datasource_b in
  issue_detail
• Advisories with empty summaries are excluded

Migration (0117_add_alias_and_summary_issue_types.py)

• Updates issue_type field choices on both AdvisoryToDo and
  AdvisoryToDoV2

Tests (test_compute_advisory_todo_v2.py) — 5 new tests

• test_relate_advisories_by_aliases_creates_todo
• test_relate_advisories_by_aliases_same_datasource_not_flagged
• test_detect_similar_summaries_creates_todo
• test_detect_similar_summaries_below_threshold_not_flagged
• test_detect_similar_summaries_empty_summary_skipped

Approach

Both new steps follow the exact same pattern as the existing
detect_conflicting_advisories() step — using LoopProgress,
bulk_create_with_m2m(), and advisories_checksum() from the
existing codebase.

The SUMMARY_SIMILARITY_THRESHOLD = 0.8 constant is defined at the
top of the file for easy adjustment.