Implement Tracker Memory Management

[plutopulp]

Problem

DownloadTracker._downloads dictionary grows unbounded. Completed downloads remain in memory forever, causing memory leak in long-running applications.

Current behavior:

# tracker.py - downloads never removed
self._downloads[url] = DownloadInfo(...)  # Stays forever

Impact on long-running apps:

• 100,000 downloads × ~500 bytes/info = ~50MB minimum
• Grows indefinitely with no cleanup mechanism
• Speed metrics are cleared, but DownloadInfo persists

Proposed Solutions

Option A: Manual clear method (simplest, non-breaking)

def clear_completed(self, max_age: timedelta | None = None) -> int:
    """Remove completed downloads from memory.
    
    Args:
        max_age: Only clear downloads completed before this duration ago.
                 None = clear all completed.
    
    Returns:
        Number of downloads cleared
    """
    # Implementation

Option B: Automatic cleanup with retention policy

class DownloadTracker:
    def __init__(
        self,
        retention_policy: RetentionPolicy | None = None,  # New param
        logger: "loguru.Logger" = get_logger(__name__),
    ):
        # Auto-cleanup completed downloads after N hours

Option C: Bounded cache with LRU eviction
Use functools.lru_cache pattern for download info storage.

Leaning towards

Option A for 0.x releases, then consider Option B for 1.0 if usage patterns warrant it.

Tasks

• Implement clear_completed() method
• Add clear_failed() method (optional)
• Add tests for clearing logic
• Document memory management in README
• Add example in docs showing periodic cleanup

Priority

Medium - Not critical for 1.0 if documented, but should be addressed.

Labels

enhancement, memory, medium-priority