[opt](partial update) Select publish-conflict read strategy by update…

[foxtail463]

Problem Summary:
FixedReadPlan previously preferred the row store whenever a complete row-store column was available, regardless of the caller’s update mode. This caused fixed partial-update conflict reconstruction to materialize and decode the complete JSONB row when only the updated columns were required. However, globally switching to physical-column reads would break UPSERT semantics for values such as raw Variant data. Physical-column reads also repeated segment setup for each requested column.

Solution:
Introduce an explicit read strategy selected according to the update mode. Fixed partial updates read their current projection from physical columns, while UPSERT and historical, flexible, sequence, and row-binlog paths retain row-store-first behavior. Batch physical-column reads so each planned segment is loaded once and column iterators are created and released sequentially.

[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?

[foxtail463]

run buildall