fix: restore live HKL calculation, throttle updates, and drop repeated log errors

[Osayi-ANL]

Summary

Live HKL calculation works again. Each profile's explicit HKL.SPEC energy and UB channels now win over the generated defaults, so SPEC and Bluesky configurations use their own PVs, and hkl_setup no longer requests the removed motor name fields that stopped geometry setup. A NaN motor position PV now fails setup with a logged error naming the PV, instead of silently turning every H, K, L value into NaN.

HKL updates are throttled to PREVIEW['HKL_UPDATE_INTERVAL_MS'] (500 ms) instead of a fixed 100 ms.

The HKL calculation reruns on every HKL PV change, so one failure would write the same error to the log thousands of times. LogManager now takes repeat_window (keyword, default 60 s) and get_logger(..., drop_repeats=True) writes a repeated message once per window. BaseWindow viewers opt in with log_drop_repeats; the Area Detector does, and its HKL messages go through its logger instead of print().

Commits

1. fix: restore live HKL calculation
2. fix: throttle live HKL updates
3. add: LogManager repeat filter

Validation

pytest tests/unit (872 passed, each commit passes on its own), ruff check src/dashpva/