search_phase_execution_exception thrown when parent and child have different id types

[dealertouch]

PGSync version: main branch

Postgres version: latest

Elasticsearch/OpenSearch version: latest

Redis version: latest

Python version: 3.12.5

Problem Description:
Assume we have three tables, inventory (primary key is of type UUID), feature_map and features(Primary key is of type Long)
when we try the update/insert following exception is thrown

opensearchpy.exceptions.RequestError: RequestError(400, 'search_phase_execution_exception', 'failed to create query: For input string: "6cc0eb22-2703-4fce-bf24-eafe6900f059"')

This is due to _meta.features.id being of type long in elasticsearch but pgsync trying to create a terms query with value "6cc0eb22-2703-4fce-bf24-eafe6900f059"

After doing some investigation, the reason behind this failure is this peace of code in symc.py:_root_foreign_key_resolver() function

        foreign_values: list = [
            payload.new.get(key) for key in foreign_keys[node.name]
        ]

this code will generate a terms query like [1, '6cc0eb22-2703-4fce-bf24-eafe6900f059'] which is invalid (two different types)
there are two issues with this code, first is the one I pointed out in this ticket with different id types, the other one I believe is since it's adding both foreign key values in feature_map table, even if they are the same type it's looking for the wrong id since it's inserting id for inventory too

I was able to fix this with replacing this with the following code (Column is of sqlalchemy type)

        foreign_values: list = []
        for column in node.columns:
            if isinstance(column, Column):
                for foreign_key in column.foreign_keys:
                    referenced_column = foreign_key.column
                    base_table = referenced_column.table
                    if base_table.name == node.parent.table:
                        foreign_values.append(payload.new.get(column.name))

also, in general, when there is an exception is thrown, the pgsync completely stops processing any data instead of moving to the next data