remote(…, s3Cluster(…)) forces distributed fragments to log is_initial_query = 1

[filimonov]

Describe the bug
Introduced in PR #583 via special handling in src/QueryPipeline/RemoteQueryExecutor.cpp:412

To Reproduce

• Repro: run SELECT * FROM remote('localhost:9000', s3Cluster(...)) and then check system.query_log
• Expected: initiator entry has is_initial_query = 1, subqueries spawned on workers log is_initial_query = 0.
• Actual: every entry in system.query_log shows is_initial_query = 1, breaking diagnostics

Additional context

Maybe we can consider using collaborate_with_initiator instead?