storage: export file deletion timeseries

File deletion can cause spikes in IO as the filesystem reclaims used blocks. To avoid IO spikes from impacting foreground traffic, we pace file deletions. This smoothes deletion IO over time. Smoothing is a delicate balance, because falling behind of file deletions can also impact foreground traffic.

We should export timeseries around the count and size of deleted files so that we better understand the rate at which file deletions are occurring practice, and their potential impact on foreground traffic.

May have helped in cockroachlabs/support#2673.

Jira issue: CRDB-32656