Fix type hint

[daikikatsuragawa]

Signed-off-by: Daiki Katsuragawa [email protected]

I read that the return type of _estimate_policy_gradient is not float but torch.Tensor (array-like).

I checked with mypy in an experiment.

mypy --ignore-missing-imports --disallow-untyped-def --warn-unreachable obp

Then the following error occurs.

obp/policy/offline_continuous.py:481: error: "float" has no attribute "mean"  [attr-defined]
obp/policy/offline_continuous.py:504: error: "float" has no attribute "mean"  [attr-defined]

This means that float is specified as the return type of _estimate_policy_gradient, but float has no method named mean.