feat: add opt-in progress callback to GetFlashMD5

[jgangemi]

Summary

Adds an opt-in ProgressFunc to GetFlashMD5, completing progress-callback coverage for the long-running flash operations (flash, erase, and read already report progress).

Like whole-chip erase, an on-device MD5 is a single blocking command — the device stays silent until the digest is ready, so there is no per-chunk protocol seam for byte-accurate progress. This reports a synthetic ETA (elapsed / estimated-total milliseconds) ticked against the size-scaled MD5 timeout, matching the ProgressFunc convention already used by EraseFlash/EraseRegion.

Behavior

• progress == nil → unchanged; no goroutine is spawned (opt-in cost only).
• non-nil → a ticker (500ms, monotonic, capped at estimate−1) runs while the blocking MD5 command executes; it is always stopped and joined before returning (no goroutine leak), never touches the connection, and emits a final (est, est) only on success.
• Intermediate ticks are best-effort: a panicking callback on an intermediate tick is recovered and dropped rather than crashing the process (the tick runs off the caller's goroutine); the final synchronous tick propagates, consistent with the other ProgressFunc call sites.

Tests

• nil path leaves the goroutine count unchanged; non-nil path fires at least one intermediate tick plus a final (est, est), monotonic and capped; the error path omits the final tick; the ticker joins on success, error, and panic; and an intermediate-callback panic is dropped without crashing.

[deadprogram]

More progress! 😄

Thanks for the addition @jgangemi now merging.