Buggify the behavior of full disk

Simulating the fullness of a disk (i.e., the percentage or size of the free space of a disk) can help better understand the cluster's behavior under disk space pressure.

This provides two benefits:
1) Help understand the minimum percentage (or size) of the free disk an FDB cluster can tolerate so that more disk space can be used. This can help reduce the disk cost.
2) Help understand the cluster's behavior when the disk is full. We should have enough confidence that the cluster can get out of the full disk problem in a timely manner if system operators perform certain operations (e.g., changing to a larger disk). This can help improve the system's robustness.

The issue names no files, tests, or entry points. Start by locating disk-space monitoring and any existing failure-injection or simulation entry points, then determine how disk fullness should be represented and observed across a cluster. Done means the behavior under configured full-disk conditions and recovery after operator action can be evaluated reliably.

Written by the indexing model from the issue text.