high(storage): ResilientBackend.WriteReader retries with consumed io.Reader

[xe-nvdk]

resilient.go:126-168 — silent data corruption on retry. Currently latent (zero callers). Fix: accept io.ReadSeeker.

[xe-nvdk]

Closing as obsolete: the code this describes no longer exists.

internal/storage/resilient.go was deleted in #561 (merged). ResilientBackend was dead code — every reference lived inside the file itself, with zero call sites in cmd/ or internal/, no tests, and no type assertions. It had a single commit dating to the original Go migration and was never wired in. It had also drifted behind the Backend interface (missing ReadToAt, StatFile, Type, ConfigJSON), which is what surfaced it in #320.

This issue's own text notes the bug was "currently latent (zero callers)" — that is precisely why the file was removed rather than repaired: an unused, untested wrapper would drift again at the next interface change.

The underlying observation remains valid as guidance for any future retry wrapper: retrying with a consumed io.Reader silently truncates, and retrying ReadTo duplicates data into a partially-written writer. Anyone reintroducing retry/circuit-breaking around cloud storage should take an io.ReadSeeker (or re-open per attempt) and not retry ReadTo at all. Worth writing against the interface as it stands then, and actually wiring it in.