ENH: Add cumulative integration function.

[DWesl]

Description Of Changes

Add a unit-aware cumulative integration function.
Won't be as fast as the xarray cumulative integration function,
but I'll be less likely to forget the units.

Checklist

• Closes #xxxx
• Tests added
• Fully documented

[dopplershift]

Initial thoughts:

• If we're adding cumulative_integrate, we should also have an integrate
• Unsure if something so generic is really a good fit for MetPy, but at the same time we are at the nexus of providing calcs that handle units on xarray
• Might be easier to make this a thin wrapper around cumulative_trapezoid from scipy.integrate? Tossing a preprocess_and_wrap decorator on the function should do all the xarray work, and then just need to deal with units as appropriate internally.

[DWesl]

Initial thoughts:

• If we're adding cumulative_integrate, we should also have an integrate

I wanted to get this working before trying for integrate, which would be a simple change. I had thought I was making a PR on my own fork, so I could get test runs from a machine that doesn't hang somewhere in collection.

• Unsure if something so generic is really a good fit for MetPy, but at the same time we are at the nexus of providing calcs that handle units on xarray

If it helps, it's kind of an inverse of mpcalc.first_derivative. On the other hand, XArray already has a cumulative_integrate method, and wrapping that to deal with the units would be simple.

EDIT: Just noticed np.trapz(pint.Quantity, pint.Quantity) returns a Quantity, which simplifies things. scipy.integrate.cumulative_trapezoid does not. I'm not sure if SciPy uses the NumPy machinery to notice a wrapper.

• Might be easier to make this a thin wrapper around cumulative_trapezoid from scipy.integrate? Tossing a preprocess_and_wrap decorator on the function should do all the xarray work, and then just need to deal with units as appropriate internally.

If preprocess_and_wrap is public, another option would be a gallery example using that to calculate PWAT or liquid water path with units.

[DWesl]

Suggestions to fix NameError in tests and lint errors for test docstrings.

[DWesl]

Initial thoughts:

• Unsure if something so generic is really a good fit for MetPy, but at the same time we are at the nexus of providing calcs that handle units on xarray

Looks like cf-xarray would also be an option, since they already have a differentiate:
https://cf-xarray.readthedocs.io/en/latest/generated/xarray.Dataset.cf.differentiate.html

• Might be easier to make this a thin wrapper around cumulative_trapezoid from scipy.integrate? Tossing a preprocess_and_wrap decorator on the function should do all the xarray work, and then just need to deal with units as appropriate internally.

I think I have that working now.

• If we're adding cumulative_integrate, we should also have an integrate

The preprocess_and_wrap decorator doesn't like the result of the function having fewer dimensions than the input to the function. I'm not entirely sure how to handle that.

[DWesl]

pint-xarray handles this now:

import xarray as xr, pint_xarray
ds = xr.open_dataset(...)
integral = ds.pint.quantify().integrate(...).pint.dequantify("~C")

I'll check cumulative_integrate in a bit