ESQL: Add a random sample command

[bpintea]

This adds a new command, RANDOM_SAMPLE <probability> <seed>. The command is producing at output a random selection of the rows it received on the input.

It receives a probability as a first mandatory parameter, a double in the (0.0, 1.0) interval, and an optional seed, an integer value. The former determines the selectivity of the sampling, the latter its reproducibility. If no seed is provided, a random one will be used.

The command will be "pushed down" to Lucene whenever possible, using the existing RandomSamplingQuery (used by the Random sampler aggregation).
This is as usually the most efficient way to execute the command since this controls which documents will be considered for loading (or further filtering) and it will also run the selection over an entire Lucene segment.

When no "push down" is possible, an ESQL operator caches a number of input pages up to a (minimum) number of rows
(currently equal to the configured maximum result set size) and then samples this batch. Meaning that (1) all the cached pages are first collected in memory (i.e. high memory pressure) and (2), the sampling runs over a significantly reduced document ID space, compared to that of a segment's (i.e. poor sampling with low probabilities).

[elasticsearchmachine]

Hi @bpintea, I've created a changelog YAML for you.