Add a new function to compute monthly averages

[alperaltuntas]

Checklist

• Enable and install pre-commit to ensure style-guides and code checks are followed.
• Target master for bugfixes and doc changes.
• Target devel for new features or functionality changes.
• Include documentation when adding new features.
• Include new tests or update existing tests when applicable.

Fixes: #55

Adds a new function to compute monthly averages from a given dataset that has more frequent data, e.g., 5 day means.

Summary of what the function does:

• First, the input dataset gets "grouped by" time_bound months. (The time_bound gets expanded, i.e., gets reshaped to be 1-dimensional for the purpose of grouping. All the DataArrays in the dataset are reshaped accordingly.)
• Then, a local function "weighted_monthly_mean" is applied to each group: Weights get computed for each chunk (e.g. 5-day) of each group (month), by taking into account how much of the chunk falls within the group.
• Finally, time and time_bound get corrected for each group (month).

Note 1: Parts of the function may be (unnecessarily) complicated, so a code review may perhaps be helpful.
Note 2: The existing monthly climatology function (compute_mon_climatology) may be rewritten to call this new function (compute_mon_averages) to compute the averages, and then to compute the correct climatology (pretty easily).
Note 3: This is a draft pull request for now, since I haven't tested it thoroughly. Let me know any issues you notice/encounter.

To test this function, you may run the following on cheyenne:

from glob import glob
import xarray as xr
import esmlab

files = sorted(glob('/glade/scratch/altuntas/archive/g.e20.GIAF.T62_g37.test_d2m.001/ocn/hist/g.e20.GIAF.T62_g37.test_d2m.001.pop.h.0001-1*.nc'))
ds = xr.open_mfdataset(files, decode_times=False, decode_coords=False)
ds_monclim = esmlab.climatology.compute_mon_averages(ds)
ds_monclim.to_netcdf("test.nc")

[andersy005]

@alperaltuntas, @matt-long, as a preliminary review comment, I was wondering if it would be worth exploring the usage of resample with cftime index as it's been implemented in pydata/xarray#2593?

I will add more comments for the rest of the PR tomorrow morning.

[matt-long]

@alperaltuntas, have you looked into @andersy005's suggestion? I think the new resample capability on CFTimeIndex could greatly simplify this type of application.

[alperaltuntas]

@matt-long, @andersy005, I'll look into it. Thanks.

[matt-long]

One think to keep in mind is an ability to correctly handled missing values that vary in time.

[andersy005]

@alperaltuntas, I am sorry I deleted the devel without making sure that there were no open pull requests. When you get time, please open a PR against the master branch