[opt](exec) Skip null maps for predicate batches without nulls

[Mryange]

Nullable predicate columns always carry their null maps through predicate evaluation and selector copying, even when a scanned batch is known to contain no NULL values. This adds unnecessary work to a common scan path. This change tracks each predicate column's batch null state, evaluates proven no-NULL batches through the nested column, and rebuilds an all-zero null map only when producing nullable output. Unknown or NULL-containing batches continue to use the existing nullable path.

Release note

None

Check List (For Author)

• Test

  • Regression test
  • Unit Test
  • Manual test (add detailed scripts or steps below)
  • No need to test or manual test. Explain why:
    • This is a refactor/code format and no logic has been changed.
    • Previous test can cover this change.
    • No code files have been changed.
    • Other reason

• Behavior changed:

  • No.
  • Yes.

• Does this need documentation?

  • No.
  • Yes.

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?

[Mryange]

/review

[github-actions]

Review status: changes requested for one test-coverage gap. After two full review rounds, no production correctness defect was found in the null-state optimization itself, but the added tests bypass the production state producer and therefore do not prove the new fast path or its NULL-containing fallback.

Critical checkpoints:

• Goal, scope, and performance: the change is small and focused, and the implementation avoids nullable predicate/null-map work only after a batch is proven to contain no top-level NULLs. Conservative UNKNOWN/HAS_NULLS paths preserve existing behavior.
• Data correctness and parallel paths: continuous and chunked reads, rowid fallbacks, index/count placeholders, scalar/default/virtual/complex/Variant iterators, dictionary conversion, vectorized and short-circuit predicates, selector copying, casts, limits, reverse reads, and lazy materialization were traced. No reachable false NO_NULLS, stale state, dropped NULL, or column-size mismatch was found.
• Concurrency and lifecycle: SegmentIterator state is thread-confined, initialized once, and reset for every predicate-column batch. No new locking, deadlock, static-initialization, ownership, or release concern applies.
• Configuration and compatibility: no configuration, FE-BE protocol, storage format, persistence, transaction, data-write, or rolling-upgrade behavior changes.
• Error handling and observability: existing Status propagation and scan/predicate timers remain appropriate; no new metric or log is required for this local hot path.
• Testing: the new helper tests are registered, but the commit states they were not run locally and all three inject PredicateColumnNullState directly. The live macOS check failed earlier in unrelated CMake/base integration with MAKE_TEST=OFF, so it neither compiled nor ran this test. A production-path nullable SegmentIterator test is still required.
• User focus: no additional review focus was provided.

Review completion: converged in Round 2. Both normal full-review reruns and the separate risk-focused audit returned NO_NEW_VALUABLE_FINDINGS with the accepted comment duplicate-fenced.

[github-actions]

Please exercise the production state producer in this test file. All three tests assign _predicate_column_null_states directly, so a regression in _read_columns_by_index—for example, a NULL-containing later chunk incorrectly remaining NO_NULLS, or every real no-NULL batch remaining conservative—would still pass even though that state controls predicate evaluation and null-map elision. Add a SegmentIterator test backed by real nullable segment data that covers no-NULL and NULL-containing batches plus continuous and rowid-fallback/discontinuous reads, and asserts returned rows/null maps for representative vectorized and short-circuit predicates.

实际跑p0就行了，beut比较难模拟实际

[Mryange]

run buildall