[roll] Roll infra/recipes [build] Record disk usage telemetry across build architectures

Add a dedicated `disk_util` recipe module and integrate disk usage
telemetry into the `fuchsia/build` recipe to inform fleet-wide Hyperdisk
and persistent disk sizing.

- Automatically detects and measures the target path across all cache
  architectures: shared cache (/mnt/shared-cache), swarming incremental
  cache ([CACHE]/incremental), and ephemeral clean builds.
- Queries total and used filesystem space (statvfs) alongside a
  best-effort build directory size (du with a 30-second timeout).
- Emits output properties `disk_used_gb` (filesystem usage for ENOSPC
  headroom monitoring), `build_size_gb` (build tree footprint for
  Hyperdisk migration sizing), and `disk_cache_architecture`.
- Executes unconditionally in the `finally:` block so telemetry is
  collected for both successful and failing builds.
- Guaranteed non-fatal: subprocess failures and parsing errors mark the
  step as a WARNING without failing or blocking the build.

Original-Original-Bug: b/482892823
Original-Original-Reviewed-on: https://fuchsia-review.googlesource.com/c/infra/recipes/+/1806877
Original-Original-Revision: 17e785f59021093299a3b5d87eb9f5c1260a1130
GitOrigin-RevId: 4e5b076945f4c764a0c76225b6db3de3d48b599c
Change-Id: Ia944503f1e2297d7e14695f66546e2009addf961
Reviewed-on: https://fuchsia-review.googlesource.com/c/integration/+/1803245
Cr-Commit-Position: refs/heads/main@{#201343}