feat: ApproxPercentileCont supports sketches from data source

[jychen7]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
Currently, approx_percentile_cont(column, quantile) (from #1538) supports raw data as input and build sketches during query time.
In the scenario of low latency query OLAP system (e.g. Druid), one common way is to pre-aggregate sketches during ingestion time (e.g. Spark/Flink -> DataStore), then merge sketches in query time (e.g. DataStore -> Datafusion).

Describe the solution you'd like
Improve approx_percentile_cont(column, quantile) to accept weight, since the pre-aggregate TDigest is just an array of (median, weight). Seems we can not change the arguments of approx_percentile_cont for backward compability, we can add new aggregator called approx_percentile_cont_with_weight

Trino have implement similar function, approx_percentile(x, percentage) and approx_percentile(x, w, percentage)

Describe alternatives you've considered
Improve approx_quantile(column, quantile) to accept an optional 3rd params, e.g. approx_quantile(column, quantile, format) where format can be

• raw (default)
• tdigest_base64
• etc (for future sketch algo, e.g KLLSketch, MomentSketch, DDSketch)

Here is an example tsv file from Druid which also use TDigest algorithm. The data contains ["timestamp", "product", "sketch"]" and it is encoded using TDigest Verbose mode

However, I eventually figure out this approach is not elegant, because it relies on the store to be encoding with TDigest's java implementation.

[jychen7]

Improve approx_quantile(column, quantile) to accept an optional 3rd params, e.g. approx_quantile(column, quantile, format)

I have implement approx_percentile_cont_from_sketch at my fork repo approx_percentile_cont_from_sketch.rs with test case tests/sql/aggregates.rs and similar csv sketch file as Druid, but I find it is not elegant enough.

So want to try another way to introduce approx_percentile_cont_with_weight(column, weight_column, percentile) similar to Trino approx_percentile(x, w, percentage)