fix: resolve flight connection leak

[dantengsky]

I hereby agree to the terms of the CLA available at: https://databend.rs/dev/policies/cla/

Summary

Execution of distributed tasks may result in a leak of Flight TCP connections.

---

How to reproduce:

• 3 node cluster (local fs storage)

• set up test data create table t as select * from numbers(10000000)

• execute the following script several times (typically, 10 times is sufficent)

bendsql  --dsn "databend://root:@localhost:8000/?sslmode=disable&enable_distributed_replace_into=1" \
         --query="replace into t on(number) select * from t"

---

turns out the primary reason for the Flight RPC connection leak is the mismatch between runtimes:

the runtime used to create the FlightClient and the one used to drop it are not the same.

Note: I cannot confirm that this completely resolves all issues related to Flight connection leaks.

---

This PR that havs been tested through the following scenarios:

• The test scenario of fix: killing in-progress tasks causes query node to crash (Segment Fault) #13847 (comment)
  conducted over approximately 12 hours, ec2 + s3, 3 node cluster
• https://github.com/datafuse-extras/test-scripts/tree/main/replace-into

These tests show that the number of open file descriptors is under controlled.

• Closes #issue

---

This change is 

[github-actions]

Docker Image for PR

• tag: pr-13866-e4c3c76

note: this image tag is only available for internal use,
please check the internal doc for more details.

[github-actions]

ClickBench Report

• hits: https://benchmark.databend.rs/clickbench/pr/13866/7049070702/hits.html
• tpch: https://benchmark.databend.rs/clickbench/pr/13866/7049070702/tpch.html

[dantengsky]

@zhang2014 PTAL, thanks