FindDiskKiller 1.2.3

FindDiskKiller 1.2.3

This patch release makes disk-write attribution and application history easier
to interpret, while fixing several layout issues across supported languages.

Improvements

• Show physical-device writes, writes attributed to visible processes, the
  unattributed remainder, and process coverage in the same sampling window.
• Keep stable application identities in daily history so continuously active,
  lower-throughput applications are not displaced by later activity spikes.
• Label the bounded historical fallback as an aggregate of other applications
  and explain that its already-aggregated legacy values cannot be split back
  into individual applications.

Fixes

• Prevent temporary executable paths from creating a new history identity on
  every build or launch, and merge compatible legacy identities during the
  schema migration without changing the stored totals.
• Fix the Apps page shifting downward after search by using a stable, explicit
  search field and responsive toolbar layout.
• Prevent long localized labels from wrapping into misaligned controls in
  narrow windows.
• Require signed Sparkle release notes so manual update checks can validate the
  complete update response.

Accuracy and safety

• Process attribution includes all visible process samples, not only the rows
  displayed in the application table. Short-lived processes, kernel I/O, and
  filesystem writeback remain explicitly unattributed rather than being
  guessed onto an application.
• History identity retention remains bounded by the existing database budget
  and retention policy; this release does not add unbounded disk usage.
• Existing AI Storage safeguards remain unchanged: analysis never starts
  automatically, scanner database access remains read-only, and cleanup uses
  only supported provider capabilities.