Refactor indirection analysis helpers

[pixge]

Motivation

• The IndirectionAnalyzer.analyze_file function had many conditionals and responsibilities, making it hard to read and maintain.
• Breaking the logic into smaller helpers reduces per-unit complexity and centralizes alias/call resolution behavior.
• Helper methods make it easier to reason about self-call resolution, alias caching, and ignorable call checks.

Description

• Split the main analysis flow into focused helpers: _collect_builtin_names, _collect_stdlib_modules, and _collect_unit_calls to encapsulate environment and call collection steps.
• Introduced _build_results, _class_context, _resolve_unit_calls, _resolve_self_call, and _is_ignorable_call to handle per-unit call resolution and result construction.
• Kept existing alias caching and call_graph computation via the existing _resolve_alias and _collect_call_graph logic, and replaced inline logic with calls to the new helpers.
• Regenerated report.json to reflect updated metrics after the refactor.

Testing

• Ran the CLI analysis command python3 -m codra.cli.main /workspace/codra --output /workspace/codra/report.json and it completed successfully with the report written to /workspace/codra/report.json.
• No automated unit tests were modified or explicitly run as part of this change.

---

Codex Task