[AI-FSSDK] [FSSDK-12750] Use attribute id instead of key for CMAB prediction requests

[raju-opti]

Summary

The CMAB service's filter_attributes method was using the attribute's key as the map key when building the filtered attributes hash for CMAB prediction requests. The CMAB prediction endpoint expects the attribute's id in the "id" field of the JSON payload, not the key. This change updates the method to use the attribute id as the hash key so the downstream HTTP request sends the correct identifier.

Changes

• Updated filter_attributes in CmabService to use attribute_id instead of attribute_key as the hash key for filtered attributes
• Added a defensive guard for malformed datafiles where attributeIds is nil or non-Array
• Updated existing test expectations to reflect attribute IDs as hash keys and added edge case tests

Jira Ticket

FSSDK-12750