fix(heal): preserve timeout budget across retries

[marshawcoco]

Related Issues

Related to #5950.

Summary of Changes

• Carry the remaining configured execution timeout into each recoverable heal retry instead of resetting the full timeout for every attempt.
• Treat an exhausted task timeout as terminal so it cannot be requeued with a fresh budget.
• Preserve the existing request and serialization shape while adding regression coverage for successive retries and manager execution.

Verification

• CARGO_TARGET_DIR=/data/rustfs-5950-pr-target /root/.local/bin/rustfs-cargo-safe test -p rustfs-heal --lib retry_request -- --nocapture --test-threads=1
• CARGO_TARGET_DIR=/data/rustfs-5950-pr-target /root/.local/bin/rustfs-cargo-safe test -p rustfs-heal --lib timeout -- --nocapture --test-threads=1
• CARGO_TARGET_DIR=/data/rustfs-5950-pr-target /root/.local/bin/rustfs-cargo-safe clippy -p rustfs-heal --all-targets -- -D warnings
• /root/.local/bin/rustfs-cargo-safe fmt --all --check

Impact

Long-running heal operations now have one aggregate execution-time budget across recoverable retries. Queue wait semantics, configuration names, defaults, and wire formats are unchanged.

Additional Notes

N/A

[github-actions]

CLA requirements are satisfied for this pull request.