Build a small `lru_estimator(cache_size, total_keys, hot_ratio)` returning estimated hit ratio (float, 4 decimal places): Model: - Hot keys = `int(total_keys * hot_ratio)` — fit entirely in cache if `hot_keys <= cache_size`. - If `hot_keys <= cache_size`: hit ratio ≈ `hot_ratio` (every hot access hits, cold misses). - If `hot_keys > cache_size`: cache thrashes; effective hit ratio ≈ `(cache_size / hot_keys) * hot_ratio`. Return value rounded to 4 decimals via `round(x, 4)`.