Stabilization of results

Problem/Motivation

After deploying `redirect_audit` in production environments with complex redirect structures, some inconsistencies have been observed across the different layers of the system (detection, storage, fix workflow, and UI).

The main observed behaviors are:

- On high-volume sites, the audit may fail to complete within memory or time constraints.
- Some redirect chains can be recorded multiple times, affecting dashboard metrics.
- Self-loops may be represented inconsistently between processing and UI display.

Overall, these behaviors can lead to results that do not fully align with the actual redirect topology in large-scale scenarios.

Motivation / Context

The goal of the `redirect_audit` module is to help validate redirect health and provide reliable insights for fixing issues.

In real-world environments with large and complex redirect graphs, some inconsistencies have been identified between the reported results and the actual state after analysis and correction workflows. This can reduce clarity when interpreting audit results in complex cases.

The objective of this work is to stabilize the behavior across all layers so that a full audit and fix cycle produces consistent results, operates within expected resource limits, and accurately reflects the final state in the UI.

This change is strictly focused on bug fixing and minimal hardening of existing flows. It does not introduce new features, configuration changes, or structural refactors outside the affected areas.