[FEATURE] Support various imputation methods

[kaituo]

Is your feature request related to a problem?
In the first draft, plan to support:

• Zero Imputation (ZERO): This method replaces all missing values with 0's. It's a simple approach, but it may introduce bias if the data is not centered around zero.
• Fixed Values Imputation (FIXED_VALUES): This method replaces missing values with a predefined set of values. The values are the same for each input dimension, and they need to be specified by the user.
• Previous Value Imputation (PREVIOUS): This method replaces missing values with the last known value in the respective input dimension. It's a commonly used method for time series data, where temporal continuity is expected.
• Linear Interpolation (LINEAR): This method estimates missing values by interpolating linearly between known values in the respective input dimension. This method assumes that the data follows a linear trend.

What alternatives have you considered?
Use rcf imputation methods when it is ready.