Infinite loop in LOCF when whole feature missing

[WenjieDu]

1. System Info

PyPOTS v0.2.1, not related to OS

2. Information

• The official example scripts
• My own created scripts

3. Reproduction

Given time series with a whole feature missing, current LOCF with first_step_imputation='backward', the imputer will be in an infinite loop because carrying values forward or backward cannot fill the feature that is whole missing

4. Expected behavior

LOCF falls in an infinite loop.