feat: Support Spark Expression Decode

[YutaLin]

Which issue does this PR close?

Closes #3184

Rationale for this change

Decode is RuntimeReplaceable in Spark. It's computed through Decode.createExpr so

• 2-arg decode is transformed into StringDecode, which Comet already handles through stringDecode(utf-8 only)
• n-arg decode is transformed into CaseWhen, which Comet also supports.

We only need to add test to verify the behavior.

What changes are included in this PR?

Add decode.sql

How are these changes tested?

run encode.sql test in spark 3.4/3.5/4.0

[comphead]

Thanks @YutaLin please add also decode to docs/source/contributor-guide/spark_expressions_support.md

[YutaLin]

Hi @comphead, thanks for the review, i've add decode to spark_expressions_support

[YutaLin]

Hi @parthchandra
Can you help me trigger CI? Thanks!

[comphead]

Thanks @YutaLin. I'm not sure why we need oracle style? it might be confusing, other than that the PR looks good to go

[YutaLin]

Hi @comphead, thanks for the review!
This is the legacy sql function originally created by Oracle. Before standard sql introduced case when statement, Oracle created Decode as a quick way to write If-Then-ELSE logic inside a query. Spark supports it purely for backward compatibility.

-- Oracle-style DECODE
SELECT DECODE(status_code, 1, 'Active', 2, 'Suspended', 'Unknown') FROM users;

-- Standard SQL Equivalent
SELECT 
  CASE status_code 
    WHEN 1 THEN 'Active'
    WHEN 2 THEN 'Suspended'
    ELSE 'Unknown'
  END
FROM users;

[comphead]

Ic, lets just keep the test case without mentioning Oracle

[comphead]

I took the liberty to update the test, once CI is done we can merge the PR, thanks @YutaLin