Update benchmark field mapping

[vmilosevic]

Updated according to comments:

I briefly looked at the data and I'd like to give some suggestions to improve data quality in the future:
config_params is Optional[dict], you can leave it as null, instead of putting an empty dict
Even though input_sequence_length / output_sequence_length / image_dimension are optional fields, I strongly recommend you to fill in some of them depending on the model type you're benchmarking (LLM vs. CV models). This will benefit any downstream analysis utilizing benchmark data that could happen in the future.
run_type is a required attribute, please fill in something meaningful instead of putting an empty string to bypass the Pydantic model checks. Same suggestion for step_name inside the measurements.

[github-actions]

Coverage Report

File	Stmts	Miss	Cover	Missing
src
benchmark.py	40	3	92%	53–55
cicd.py	64	3	95%	85–87
generate_data.py	32	7	78%	48–66
pydantic_models.py	97	0	100%
test_parser.py	25	11	56%	31–36, 40–45
utils.py	100	4	96%	44, 123, 176–179
src/parsers
junit_xml_utils.py	47	6	87%	85–89, 93–97
parser.py	10	2	80%	23, 33
python_pytest_parser.py	69	6	91%	19, 52, 55, 81, 131–132
python_unittest_parser.py	61	1	98%	30
tt_torch_model_tests_parser.py	62	0	100%
TOTAL	607	43	93%

Tests	Skipped	Failures	Errors	Time
6	0 💤	0 ❌	0 🔥	0.517s ⏱️