perf: improve ScalarValue::to_array_of_size for Boolean and some null values

[rluvaton]

Which issue does this PR close?

N/A

Rationale for this change

make stuff faster

What changes are included in this PR?

1. updated Boolean scalar to array to use existing functions for creating BooleanArray with the same value n times
2. Update None Binary/BinaryView/FixedSizeBinary/LargeBinary to use new_null_array instead of repeat

Are these changes tested?

Existing tests

Are there any user-facing changes?

Nope

[alamb]

Here is a related item: #18159

The CI test failure https://github.com/apache/datafusion/actions/runs/18657662446/job/53190612662?pr=18180 looks like it may be a flaky test (there is guarantee that shuffle does not leave the input array the same)

I will restart it

running 1 test
test engines::conversion::tests::test_big_decimal_to_str ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

     Running bin/sqllogictests.rs (target/ci/deps/sqllogictests-1b770706aacdb3f8)
External error: 1 errors in file /Users/runner/work/datafusion/datafusion/datafusion/sqllogictest/test_files/spark/array/shuffle.slt

1. query result mismatch:
[SQL] SELECT shuffle([1, 2, 3, 4, 5, NULL]) != [1, 2, 3, 4, 5, NULL];
[Diff] (-expected|+actual)
-   true
+   false
at /Users/runner/work/datafusion/datafusion/datafusion/sqllogictest/test_files/spark/array/shuffle.slt:24

[alamb]

Thank you @rluvaton

[adriangb]

I wonder if this will benefit from #16789 or if we could expand that sort of approach to more places.

[rluvaton]

I wonder if this will benefit from #16789 or if we could expand that sort of approach to more places.

I'm not familiar with that PR, do you plan to replace all nulls creation?

[adriangb]

The basic idea is that the num_rows parameter is largely constant (the batch size). IIRC I only implemented caching in that PR for dictionary keys and arrays of type null but there's not reason we couldn't expand this to more data types and call sites. Ultimately any call to array_of_size where the ScalarValue is null should be able to re-use the same underlaying buffer of all null values I think...