Improve decompression logic on restoring compressed backup

[donhardman]

At present, our backup restoration process involves reading the file into memory and decompressing it within the same process. This method can lead to "memory out of limit" errors when dealing with large files. To rectify this, we should adopt a more efficient approach for decompression by leveraging the streams of the zstd extension. This will help minimize memory consumption during the restoration of substantial files.

[donhardman]

Implemented in this pull: https://github.com/manticoresoftware/manticoresearch-backup/pull/85