Add StringSplitAggregator

[gsarti]

Description

This PR introduces StringSplitAggregator (nickname for aggregate usage: "split"), supporting aggregating on token groups after splitting on arbitrary substrings.

Example usage:

# Split on newlines. Default split_mode = "single" (tokens with newlines are kept separate in the aggregated result.
out.aggregate("split", split_pattern="\n").aggregate("sum").show(do_aggregation=False)

# Split on whitespace-separated words of length 5. Split mode "end" means that tokens on which the sequence is split are concatenated to the preceding token sequence.
out.aggregate("split", split_pattern=r"\s(\w{5})(?=\s)", split_mode="end")

Example multi-example prompt with newline splitting:

split_mode = "single"

split_mode= = "start" (sequences begin with \n)

split_mode= = "end" (sequences end with \n)