Added min_merge_bytes_to_use_direct_io Option for MergeTree Engines

Why it matters

This feature allows users to specify a size threshold for data merges in MergeTree tables. When the total size of a merge exceeds this threshold, the merge process will use O_DIRECT to handle data part files, improving disk I/O efficiency and potentially reducing cache pollution during large merges.

How to use it

Set the min_merge_bytes_to_use_direct_io parameter in the MergeTree table engine settings with the desired byte threshold. For example:

CREATE TABLE example (
    ...
) ENGINE = MergeTree()
SETTINGS min_merge_bytes_to_use_direct_io = 1000000000;

This will enable direct I/O for merges larger than 1GB.