Cobalt Image Cache: Implement pre-allocated pool, logging, and statistics

[kjyoun]

• Implemented ImageDecodeMemoryPool to manage a contiguous 32MB arena for image decodes, mitigating physical memory fragmentation.
• Implemented PooledDiscardableMemory to allocate backing memory from the pool.
• Integrated the pool into GpuImageDecodeCache::DecodeImageIfNecessary, controlled via the 'PreallocatedImageCachePool' base::Feature.
• Added robust fallback to the default allocator if the pool runs out of memory.
• Added detailed logs for SW/HW decodes and cache stats in 'MESSAGE: key=value' format.
• Added explicit statistics tracking to ImageDecodeMemoryPool (allocations, failures) and integrated them into the periodic STATS log.

Issue: 1234

[github-actions]

🤖 Gemini Suggested Commit Message

---

cobalt: Add pre-allocated image decode pool

Implement a 32MB pre-allocated arena for image decodes to mitigate
physical memory fragmentation. This pool is managed by the new
ImageDecodeMemoryPool and PooledDiscardableMemory classes.

The pool is integrated into GpuImageDecodeCache and controlled via the
PreallocatedImageCachePool feature flag. This change also introduces
detailed logging for software and hardware decodes, along with periodic
statistics tracking for allocation success and failure rates to improve
observability.

Issue: 1234

---

💡 Pro Tips for a Better Commit Message:

1. Influence the Result: Want to change the output? You can write custom prompts or instructions directly in the Pull Request description. The model uses that text to generate the message.
2. Re-run the Generator: Post a comment with: /generate-commit-message