Support aggregate functions to better handle timeseries data

Some of the datasets store value of an entity (ex: humidity/temperature readings of a storage unit, Status of a device) over time. The values are recorded periodically, may not be very granular. It would be great to have some of the following timeseries functions to analyze the trend over time.

- time_bucket_gapfill() => Fills data for missing time buckets.
- Functions to fill data: zero_fill() -> Fill missing buckets with zero; locf() => Fill from previous value; interpolation.
- Extrapolation functions for time series forecasting.
- Aggregate functions to chose value for a time bucket such as first() / last() along with avg and percentile functions.

cc @elonazoulay

The issue names no files, tests, or entry points. First clarify which time-series functions and aggregation behavior are in scope, then locate Pinot's aggregate-function implementation and relevant tests; done means the selected functions support the requested time-bucket analysis with coverage for missing data and aggregation semantics.

Written by the indexing model from the issue text.