[Platform] Identify and selectively deduplicate shared code hot spots

Summary

Identify duplicated and frequently modified code paths across model-family implementations, then consolidate the parts that represent stable shared behavior.

We should preserve intentional model-family isolation and avoid creating abstractions solely to reduce line count.

Tasks

• Identify the highest-churn and most frequently duplicated code paths.
• Classify each case as intentional isolation or accidental duplication.
• Refactor the highest-value candidates.
• Add regression tests for the affected model families.
• Document which code should remain intentionally duplicated.

Acceptance criteria

• A ranked inventory of code hot spots is available.
• Selected shared code is consolidated without changing behavior.
• Existing model qualification and performance tests continue to pass.
• Remaining candidates have clearly scoped follow-up issues.

Start by reviewing the model-family implementations and identifying duplicated, frequently modified paths, then classify intentional isolation versus accidental duplication. Use the existing model qualification and performance tests as regression checks; done means a ranked inventory, selected behavior-preserving consolidations, documented intentional duplication, and scoped follow-up issues.