Faster timeseries, breakdown, and list-of-views queries

We have rolled out a set of query and storage optimizations across the metrics service. The biggest wins are on the endpoints customers hit most often from dashboards and BI integrations.

Improvements include:

• Optimized timeseries query — the timeseries path has been rewritten to push more work into Pinot, reducing the volume of data shipped back to the API tier.
• Optimized breakdown and overall queries — the breakdown, overall, and overview endpoints use a new query shape with is_final semantics that skips redundant scans of in-progress sessions.
• Optimized list-of-views and top-content — list-of-views and top-content queries have been restructured to use indexed columns and produce results faster on large catalogs.
• Optimized errors APIs — error timeseries and breakdown endpoints use the same optimized pattern, fixing both speed and correctness for total-views counts in error views.
• Pinot tuning — UTF-8 handling and table-level configuration changes mean fewer fallbacks to slow paths on multi-byte titles and metadata.

No client changes are required. Existing queries will simply respond faster, with the largest gains on long time ranges and high-cardinality breakdowns.