Fix ValueError in concatenation of result_by_req_id with zero-dimensional arrays, when there are 1-length arrays in the inference result.

[pierce-ahn]

Description

This PR addresses an issue where the concatenate function encounters zero-dimensional arrays during the processing of result_by_req_id, causing a ValueError. Specifically, the error message observed was:

ValueError: zero-dimensional arrays cannot be concatenated

Example of the problematic result_by_req_id:

result_by_req_id: [[array([b'[[0, 1000, "|\xe3\x82\x8c"], [1020, 180, "|\xe3\x81\xa1\xe3\x82\x87\xe3\x81\xa3\xe3\x81\xa8"], [1220, 180, "|\xe3\x82\xa2\xe3\x83\x97"], [1440, 40, "|\xe3\x81\xa7"], [1500, 80, "|\xe8\xa6\x8b"], [1640, 40, "|\xe3\x81\xa6"], [1700, 180, "|\xe3\x82\x82\xe3\x82\x89\xe3\x81\x86"], [1900, 240, "|\xe3\x81\xa8"]]'], dtype=object), array(0.08090031, dtype=float32)]]

Changes

• Added safe_concatenate function: This function ensures that any zero-dimensional arrays are expanded to one-dimensional arrays before concatenation.
• Updated the map function: The map function now uses safe_concatenate to handle each element of zipped_result safely.

[kimdwkimdw]

Any example model repository for this PR? Please add an example in https://github.com/rtzr/tritony/blob/main/tests/test_model_call.py.