Upgrade Pandas version to latest version >=0.23.1

[dirkesquire]

I love marbles! I also enjoyed Jane's talk at PyData Amsterdam.

My issue is that installing marbles automatically downgrades your pandas version.

If you do: pip install pandas it installs version 0.23.1 at the time of writing.
Then running pip install marbles downgrades pandas to 0.21.

This causes the following code to break:

from datetime import datetime
import pandas as pd

data = {
            datetime(2018,4,1): [2500, 1500],
            datetime(2018,5,1): [2000, 2500],
            datetime(2018,6,1): [2000, 1500],
    }
df = pd.DataFrame.from_dict(data, orient='index', columns = ['Income', 'Expense'])
print(df)

I believe the line is on marbles/marbles/mixins/setup.py line 57.

I could try and do a pull request if you wanted, but I'm not sure what all tests I would need to ensure are passing?

[leifwalsh]

Yep, this is something we should do. @thejunglejane what kind of testing do you want to do about it?

[thejunglejane]

We're currently using very little of the pandas API: we turn a sequence of date(time)s into a series and then shift the series. See assertDateTimesFrequencyEqual if you're curious.

I can review the pandas changelog and run our normal suite of tests with 0.22 and 0.23 and I think that'll be enough.