ChartDataQueryFailedError/ChartDataCacheLoadError map to 400/422 in the
synchronous chart/data endpoint - expected, client-facing validation
failures (e.g. a chart referencing columns a customer has since dropped
from the dataset), not application bugs. load_chart_data_into_cache
unconditionally re-raised every exception after reporting it via
update_job, so these got double-reported: once cleanly to the client,
and again as an unhandled Celery task exception (and Sentry ERROR).
Fixes SUPERSET-PYTHON-13JV
Co-Authored-By: Claude <noreply@anthropic.com>