[GLUTEN-11752] Fix AdaptiveSparkPlanExec accessibility in columnar write optimization

[wangyum]

What changes are proposed in this pull request?

This pr fixes the issue where Gluten's columnar writer optimization breaks shuffle IDs retrieval introduced in Spark PR #51432.

Spark expects to access AdaptiveSparkPlanExec.shuffleIds via pattern matching:

queryExecution.executedPlan match {
  case ae: AdaptiveSparkPlanExec =>
    ae.context.shuffleIds.asScala.keys
}

This PR refactored the wrapping logic to:

1. Wrap aqe.inputPlan with genColumnarToCarrierRow() first → ColumnarToCarrierRow(inputPlan)
2. Create a new AdaptiveSparkPlanExec with the wrapped child → AdaptiveSparkPlanExec(ColumnarToCarrierRow(...))
3. Set supportsColumnar=false since the child is already columnar

This preserves AdaptiveSparkPlanExec in the plan hierarchy while maintaining the columnar write optimization benefits.

How was this patch tested?

Unit test.

Was this patch authored or co-authored using generative AI tooling?

No.

Fixes #11752

[github-actions]

Run Gluten Clickhouse CI on x86

[github-actions]

Run Gluten Clickhouse CI on x86

[github-actions]

Run Gluten Clickhouse CI on x86

[pan3793]

was it fixed by apache/spark#53620?

[zhouyuan]

@wangyum could you please do a rebase to fix the CK CI issue?
Cc @JkSelf

[github-actions]

Run Gluten Clickhouse CI on x86

[wangyum]

@pan3793

1. Better compatibility - Maintaining plan structure consistent with Spark avoids similar issues with other code that depends on AdaptiveSparkPlanExec.
2. Clearer semantics - Having AdaptiveSparkPlanExec as the child with other operators wrapping it aligns better with Spark's design intent.

[github-actions]

Run Gluten Clickhouse CI on x86