Implement S3 partitioned reads

[arthurpassos]

Describe the new feature

Being able to read from S3 partitioned table.

Writes to partitioned table is already supported, but reading not. I suspect the reason is that it is not clearly defined how to determine if a file belongs to the table or not. Should it be treated as a star wildcard? Or we need to create a regex expression for the partition id?

The implementation will be useful for multi file tables

Use case

CREATE TABLE s3 (event_date DateTime64, val str) ENGINE=S3('aws-s3-bucket/{_partition_id}/file.parquet') PARTITION BY event_date SETTINGS s3_create_new_file_on_insert=1;

select * from s3;

Alternative solutions

Are there other ways to solve the problem this new feature addresses? Are any of them possible now? If so, how would the solution you're proposing be better?

Additional context

Add any other context about your proposed feature here.

[arthurpassos]

This should go to upstream