[INLONG-12125][SDK] Enhance Transform SDK protobuf processing and SQL alias parsing

[luchunliang]

Fixes #12125

Motivation

This PR introduces several enhancements to the Transform SDK for protobuf data processing and SQL alias parsing:

1. Improve PbSourceData.buildFieldValue() with Flink-compatible type conversion
   Enhance buildFieldValue() to produce values directly compatible with Flink's RowData type system:

STRING → wrap with BinaryStringData; handle existing BinaryStringData instances
INT/LONG/FLOAT/DOUBLE/BOOLEAN → validate actual types and convert from String if necessary using NumberUtils
BYTE_STRING → handle both ByteString and byte[] input; fallback to ISO_8859_1 encoding for other types
MESSAGE → delegate to buildStructData()
Other types → return as-is
2. Improve PbSourceData node value caching with IdentityHashMap
Replace HashMap with IdentityHashMap for nodeValueCache and mapNodeCache to use reference equality (==) instead of equals() when keying by DynamicMessage instances. This avoids expensive deep-equality comparisons and ensures correct identity-based lookups for protobuf messages.

3. Fix null handling in PbSourceData.findFieldNode() and findChildField()
   Change return value from empty string "" to null when a field cannot be found, making the behavior consistent across the codebase and preventing downstream logic from treating a missing field as an empty string.

4. Fix DefaultSinkData.getField() to return null for missing fields
   Change getOrDefault(fieldName, "") to get(fieldName) so that missing fields return null instead of "". This allows downstream encoders (e.g. RowDataSinkEncoder) to correctly distinguish between "field not set" and "field set to empty string".

5. Add null filtering in array/map field processing
   In getField() array processing: skip null items via if (itemValue != null) before adding to result list
   In buildMapData(): skip map entries where key or value is null via if (keyValue != null && valueValue != null)

6. Enhance findNodeValueByCache() with more precise cache-hit handling
   When a path-level cache hit is found at the last node position, apply the full node-type resolution logic (array index extraction, map key lookup) rather than returning the raw cached value directly. This ensures correct results for paths like field(0) or map_field(key) when the parent list/map was cached from a prior access.

7. Fix SQL alias parsing for special characters (backtick handling)
   In TransformProcessor.initTransformSql(), strip surrounding backtick characters from alias names returned by exprItem.getAlias().getName(). JSqlParser's getName() preserves backtick quotes in the returned string, causing a mismatch with the sink field list. This fix allows aliases containing special characters (e.g., *, $) to be properly matched when wrapped in backticks in the SQL.

Modifications

Protobuf data → Flink RowData → Iceberg pipeline requires strict type compatibility (e.g., BinaryStringData for STRING fields)
Identity-based caching avoids O(n) deep-equality for DynamicMessage keys
Consistent null semantics prevent silent data corruption when fields are missing
Aliases containing $ and * characters (common in protobuf map key paths) must be parseable in SQL

Verifying this change

(Please pick either of the following options)

• This change is a trivial rework/code cleanup without any test coverage.

• This change is already covered by existing tests, such as:
  (please describe tests)

• This change added tests and can be verified as follows:

  (example:)

  • Added integration tests for end-to-end deployment with large payloads (10MB)
  • Extended integration test for recovery after broker failure

Documentation

• Does this pull request introduce a new feature? (yes / no)
• If yes, how is the feature documented? (not applicable / docs / JavaDocs / not documented)
• If a feature is not applicable for documentation, explain why?
• If a feature is not documented yet in this PR, please create a follow-up issue for adding the documentation