[Rust][Arrow] Report request-body loss before EOF

[teodordelibasic-db]

Summary

RequestBodyControl::wait_for_eof() currently treats disappearance of the request-body EOF sender as an indefinitely pending state. The surrounding rotation drain eventually reaches its overall deadline, which hides the actual request-body failure and consumes the remaining drain budget.

Report request-body loss explicitly when the body disappears before reaching EOF.

Scope

• Change wait_for_eof() to distinguish clean EOF from loss of its signaling sender.
• Return and propagate an explicit error when the request body disappears before EOF.
• Surface the error through Arrow rotation and teardown without waiting for the overall drain deadline.
• Preserve stable precedence for a real peer or protocol error already observed during the same teardown.
• Keep the existing bounded drain behavior; this issue does not extend or configure the post-EOF drain duration.
• Update rust/NEXT_CHANGELOG.md for the observable error-model improvement.

Acceptance criteria

• Clean request EOF still completes successfully.
• Dropping the request-body producer before EOF returns an explicit error promptly.
• The dropped-body path does not remain pending until the rotation drain deadline.
• Deterministic tests cover both clean EOF and sender disappearance.
• Existing post-EOF ACK, peer-error, and drain-timeout tests continue to pass.

Relationship to other work

• [Rust] Fix Arrow Flight rotation H2 resets #649 introduced bounded request EOF and response draining during server rotation.
• This issue is independent of [Rust][Arrow] Coordinate explicit close with recovery #657: the diagnostic remains useful even if explicit close during recovery continues to use bounded abort semantics.
• Extended or configurable post-EOF ACK draining is separate follow-up work.