sysutil: fill ballast files with random data instead of zeroes

[NAVEENKUMARKR777]

Summary

resizeLargeFileNaive is used to create emergency ballast files on
platforms without fallocate support: all non-Linux platforms, and Linux
filesystems that report fallocate as unsupported (eg some ZFS
configurations, per the reproduction in #78606). It previously wrote
zero-filled buffers to reserve the requested disk space.

Some filesystems compress long runs of zeroes down to almost nothing (eg ZFS
with compression=on), so the ballast file's logical size matched the
requested size but its actual on-disk footprint did not — as demonstrated in
the original report, a 559GB ballast occupied only ~2.7MB of pool space.
This defeats the purpose of a ballast file: reserving real, reclaimable disk
space that can be freed by deleting the file when a node runs low on disk.

This exact fix (write random data instead of zeroes for the buffer-writing
fallback path) was proposed and agreed on by @knz, @nicktrav, and @jbowens
in the issue thread back in 2022, but was never implemented.

This only addresses the buffer-writing fallback path. It does not address
the separate, harder case raised later in the thread where fallocate
itself reports success without actually reserving space on some
filesystems — that needs separate investigation (eg checking on-disk usage
after allocation) and is out of scope here.

Test plan

Added TestResizeLargeFileNaiveIncompressible, which writes a ballast file
via the naive path and verifies a standard compressor (compress/flate)
cannot shrink its content by more than a small margin, directly regression
testing the compressibility issue from #78606. Existing TestResizeLargeFile
continues to pass unmodified.

Resolves: #78606
Epic: none

Release note (bug fix): Fixed a bug where the automatically-created
emergency ballast file could fail to reserve real disk space on filesystems
that compress zero-filled data (such as ZFS with compression enabled),
preventing it from being useful as a safety margin when a node ran low on
disk space.

🤖 Generated with Claude Code

[blathers-crl]

Thank you for contributing to CockroachDB. Please ensure you have followed the guidelines for creating a PR.

My owl senses detect your PR is good for review. Please keep an eye out for any test failures in CI.

I have added a few people who may be able to assist in reviewing:

• @nicktrav (commented on storage: ballast file has no effect under zfs with compression=on #78606)

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}

[cockroachlabs-cla-agent]

All committers have signed the CLA.