Feature: Interval joins

[gwaramadze]

Implement interval join

This PR introduces a new join_interval() method to StreamingDataFrame for performing stream-to-stream interval joins.

Interval Join

A new interval join feature has been added, allowing records from two streams to be joined based on a user-defined time interval around their timestamps.

Key Features:

• Symmetric Interval: Configure the join with backward_ms and forward_ms to define a flexible time window for matching records between streams.
• Join Types: Supports both inner and left joins.
• Flexible Merging: Offers built-in strategies (raise, keep-left, keep-right) and allows custom functions for merging matched records.
• Stateful Efficiency: Leverages an enhanced TimestampedStore that can now store and query multiple values per key and timestamp, which is crucial for handling multiple matches within an interval.

Example Usage:

# Join records from `left_sdf` with records from `right_sdf` that
# fall within a window of 1s before to 2s after the left record's timestamp.
joined_sdf = left_sdf.join_interval(
    right_sdf,
    how="inner",
    backward_ms=1000,
    forward_ms=2000,
)

State Store Enhancements

To support this feature, TimestampedStore has been updated with a get_interval() method to retrieve all records within a given time range.