feat: improve excel date datatype in context.preview api

[leavesster]

When pandas load data from excel, some date columns will load as datetime object.

Excel content:

date	count
2019/1/1	1
2019/1/2	1
2019/1/3	1
2019/1/4	1
2019/1/5	1
2019/1/6	1

pip install pandas openpyxl

import pandas
# from json import dumps

def main():
    df = pandas.read_excel('date.xlsx')
    print(df)
    print(df.dtypes)
    # d = dumps(df.to_dict(orient='split')) # will crash and report 'TypeError: Object of type Timestamp is not JSON serializable'

if __name__ == "__main__":
    main()

---

Currently context api use to_json instead of to_dict to avoid this error.

to_json will transform date to epoch milliseconds(default) or ISO8601(iso option) format. The default ISO 8601 format is 2019-01-02T00:00:00.000.

---

Target:

• It should be better use local timezone or what we see in Excel(the actually value is float).
• Add datetime test case