fix/perf: Thread-safe collection of semantic_skip_warnings in parallel analyze

[omsherikar]

Scope

refactron/core/refactron.py — analyze() parallel branch
refactron/core/config.py — use_multiprocessing: bool = False (threading default)
refactron/core/parallel.py — ThreadPoolExecutor

Problem

With default threading, process_file_wrapper appends to result.semantic_skip_warnings from multiple workers. CPython list append is not guaranteed safe under concurrent modification; this can cause lost updates or rare corruption under load.

Suggested direction

• Return skip warnings from process_func and merge in the main thread after process_files completes; or use a threading.Lock around appends; or use a queue.

Acceptance

• No shared mutable AnalysisResult fields mutated from worker threads; stress test optional.