Give more options in LOCF for users to impute still-missing values after LOCF running

[WenjieDu]

1. Feature description

Currently, we impute still-missing values with zeros after LOCF in the LOCF method, like the code below

PyPOTS/pypots/imputation/locf/model.py

Lines 117 to 121 in c962530

	# If there are values still missing,
	# they are missing at the beginning of the time-series sequence.
	# Impute them with self.nan
	if torch.isnan(X_imputed).any():
	X_imputed = torch.nan_to_num(X_imputed, nan=self.nan)

But this is not flexible. We need more options.

2. Motivation

Make LOCF more powerful.

3. Your contribution

Will make a PR.