fix(testing/bench.py): make_param_key use _format_value for non-width keys

[pandacooming]

Summary

Apply _format_value() to non-_WIDTH values in make_param_key(), ensuring torch.dtype and other complex types are formatted consistently with make_param_id(), which already uses _format_value.

Fixes #8

Problem

Function	Input {'dtype': torch.float16}	Output
make_param_id		dtype=fp16 ✅
make_param_key		dtype=torch.float16 ❌

Changes

def make_param_key(params: dict) -> str:
    param_str = ','.join(
-       f'{_SHORT_NAME.get(k, k)}={format(v, f">{_WIDTH.get(k)}") if k in _WIDTH else v}'
+       f'{_SHORT_NAME.get(k, k)}={_format_value(v) if k not in _WIDTH else format(v, f">{_WIDTH.get(k)}")}'
        for k, v in params.items() if v is not None
    )
    return f'{param_str}'

Also changes v != None to v is not None (PEP8 style).

Testing

make_param_key({'dtype': torch.float16, 'num_tokens': 1024, 'hidden': 256})
  → dtype=fp16,num_tokens= 1024,hidden= 256

make_param_id({'dtype': torch.float16, 'num_tokens': 1024, 'hidden': 256})
  → dtype=fp16-num_tokens=1024-hidden=256

dtype part match: True  (dtype=fp16 vs dtype=fp16) ✅