Implement sample by size on ray runner #5690
Description
Is your feature request related to a problem?
Currently sampling by size (i.e. num rows) only works on the native runner.
Describe the solution you'd like
I would like to do sampling by size on the ray runner.
Describe alternatives you've considered
No response
Additional Context
Currently, sampling by size works locally by assigning a random key to each row, and taking the minimum keys. We can extend this to distributed by gathering the local samples and then doing another round of min.
Would you like to implement a fix?
Yes