Preserve ww schema in partial dependence

[freddyaboulton]

Pull Request Description

Fixes #2928

---

After creating the pull request: in order to pass the release_notes_updated check you will need to update the "Future Release" section of docs/source/release_notes.rst to include this pull request by adding :pr:123.

[codecov]

Codecov Report

Merging #2929 (74d87f9) into main (7d97020) will increase coverage by 0.1%.
The diff coverage is 100.0%.

@@           Coverage Diff           @@
##            main   #2929     +/-   ##
=======================================
+ Coverage   99.7%   99.7%   +0.1%     
=======================================
  Files        302     302             
  Lines      28396   28412     +16     
=======================================
+ Hits       28303   28319     +16     
  Misses        93      93             

Impacted Files	Coverage Δ
evalml/model_understanding/_partial_dependence.py	98.8% <100.0%> (+0.1%)	⬆️
...del_understanding_tests/test_partial_dependence.py	99.3% <100.0%> (+0.1%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7d97020...74d87f9. Read the comment docs.

[chukarsten]

Awesome work, Freddy! Just a quick question about the copy in partial dependence, but nothing blocking.

[chukarsten]

Is there a reason we copy this first rather than just build a new df by concatting the series? I've seen this a few times but never had the courage to ask :)

[freddyaboulton]

The partial dependence computation requires us to fill a given feature will only one value while keeping all other features the same. To not override the user's original data, I think we need a copy. Since we need all the features to be present in the data, I think concatting all the features will be equivalent to a copy + modify operation (and result in the same memory).

[eccabay]

Solid, LGTM!