Bug fix for the WMA operator

[dmnsn7]

Description

A quick fix for the WMA operator.

The bug can be reproduced by simply running some code (slightly modified) from the examples/tutorial/detailed_workflow.ipynb.

In [ ]:
df = D.features(
    ["SH601216"],
    ["$open", "WMA($open,3)", "WMA($open,3)*3"],
    start_time="2020-05-01",
    end_time="2020-05-31",
)
print(df)

Out [ ]:
                          $open  WMA($open,3)  WMA($open,3)*3
instrument datetime                                          
SH601216   2020-05-06  1.300132      0.431863        1.295588
           2020-05-07  1.318316      0.436666        1.309997
           2020-05-08  1.309334      0.436931        1.310794
           2020-05-11  1.313896      0.437704        1.313112
           2020-05-12  1.309334      0.436952        1.310855
           2020-05-13  1.309334      0.436698        1.310095
           2020-05-14  1.304933      0.435711        1.307133
           2020-05-15  1.295523      0.433654        1.300961
           2020-05-18  1.291245      0.431651        1.294952
           2020-05-19  1.295807      0.431413        1.294239
           2020-05-20  1.314077      0.434727        1.304182
           2020-05-21  1.359704      0.444615        1.333846
           2020-05-22  1.350579      0.449179        1.347537
           2020-05-25  1.363965      0.452931        1.358793
           2020-05-26  1.358994      0.453083        1.359248
           2020-05-27  1.307209      0.444643        1.333930
           2020-05-28  1.276208      0.433446        1.300339
           2020-05-29  1.286643      0.428864        1.286592

Motivation and Context

Omitted.

How Has This Been Tested?

Just let it simply align with the implementation of the EMA operator.

Screenshots of Test Results (if appropriate):

Omitted.

Types of changes

• Fix bugs
• Add new feature
• Update documentation