[spark] Improve SQL UDF with body analysis and validation

[Zouxxyy]

Purpose

Follow-up to #8299. Aligns Paimon SQL UDF creation with Spark's CreateSQLFunctionCommand:

• Body analysis: move analyze/validate/derive from parser rule to CreatePaimonSQLFunctionCommand.run() (execution stage), mirroring Spark's approach
• Validation: reject aggregate/window/generator in scalar function body, duplicate parameter names, non-trailing defaults, NOT NULL parameters, temp object references, invalid default expressions
• Derivation: infer deterministic and containsSQL from the analyzed plan
• Return type inference: support omitting RETURNS clause (Spark 4.1+)
• Properties: capture SQL configs (timezone, ANSI mode) and catalog/namespace at creation time, restore at query time
• Cleanup: remove redundant options (INPUT_PARAM, RETURN_TYPE), unify DEFAULT metadata key handling in SparkTypeUtils

Tests

• SQL configs round-trip (ANSI mode preserved across sessions)
• Non-deterministic function body (rand())
• Reject aggregate/window/generator in scalar body
• Reject duplicate parameter names
• Reject non-trailing defaults
• Omitting RETURNS clause (Spark 4.1+ only)