[CalvinKirs]

What problem does this PR solve?

Related PR: #69314

Problem Summary:

• COM_STMT_SEND_LONG_DATA returns an unexpected response and does not bind uploaded chunks, leaving streaming clients out of sync. Accept chunks without replying, defer upload errors until EXECUTE, and clear uploaded data after execution or RESET.
• Reusing cached server-prepared statements through a non-master FE can fail because the master prepares each forwarded execution independently. Forward complete parameter types and values, including streamed parameters, and use the active prepare-stage context during proxy planning.
• Empty results with useCursorFetch=true can hang or leave unread packets, including when connection attributes are absent or rewritten. Return an empty cursor, complete FETCH with LAST_ROW_SENT, and preserve framing for older FEs through an optional peer capability.
• Completed EXECUTE payloads can remain attached to idle connections. Release command payload buffers after response finalization and forwarding finish.

Backport adaptations for branch-4.2:

• Keep protocol state in the existing ConnectContext/ConnectProcessor architecture and preserve branch-4.2's expansion of cached parameter types only when forwarding.
• Regenerate regression expectations with the standard test script to retain branch-4.2's existing VARBINARY function semantics.

Release note

Fix streamed prepared parameters, repeated prepared execution through non-master FEs, and empty cursor fetch hangs.

Check List (For Author)

• Test

  • Regression test: prepared_stmt_p0/send_long_data and prepared_stmt_p0/cursor_fetch_empty_result passed with run-regression-test.sh.
  • Unit Test: 91 FE tests passed with run-fe-ut.sh, with zero failures, errors or skips.
  • Manual test: Connector/J 8.0.27, 8.0.33, 8.4.0 and 9.7.0 each passed 428 checks against the built branch-4.2 FE and BE 4.1.3. Connect explicitly to master and follower endpoints; verify streamed bytes, empty streams and statement reuse; run 50 batches of three rows per version and compare all stored values; alternate empty/nonempty cursor results with and without connection attributes, including forwarded execution and connection reuse.
  • Full FE build passed via build.sh --fe; Checkstyle passed with zero violations.

• Behavior changed:

  • Yes. Long-data uploads have no response; forwarded parameters are self-contained; empty cursors complete through FETCH; completed command buffers are released.

• Does this need documentation?

  • No.

Check List (For Reviewer who merge this PR)

• Confirm the release note
• Confirm test cases
• Confirm document
• Add branch pick label

[hello-stephen]

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
3. What features were added. Why was this function added?
4. Which code was refactored and why was this part of the code refactored?
5. Which functions were optimized and what is the difference before and after the optimization?

[CalvinKirs]

run buildall

[hello-stephen]

FE Regression Coverage Report

Increment line coverage 65.31% (96/147) 🎉
Increment coverage report
Complete coverage report

[hello-stephen]

BE Regression && UT Coverage Report

Increment line coverage 100% (0/0) 🎉

Increment coverage report
Complete coverage report

Category	Coverage
Function Coverage	76.36% (34499/45177)
Line Coverage	61.17% (389486/636684)
Region Coverage	57.94% (327856/565851)
Branch Coverage	58.92% (151386/256939)

[yiguolei]

skip buildall