Switch order of boolean checks in infer_feature_types

[freddyaboulton]

Pull Request Description

Fixes #2642

I was able to track the regression down to infer_feature_types like @chukarsten and I suspected.

Steps to identify problem:

1. Profile this script locally for both v0.30.0 and main

import pandas as pd
from evalml.automl import AutoMLSearch
from evalml.preprocessing import split_data

X = pd.read_csv("/Users/freddy.boulton/Downloads/KDDCup09_churn.csv")
y = X.pop("CHURN")

X_train, X_validation, y_train, y_validation = split_data(X, y, problem_type="binary",test_size=0.5, random_seed=0)
automl = AutoMLSearch(X_train, y_train, "binary")
automl.search()

2. Visualize with snakeviz. I saw that main took 5 minutes 3 seconds while v0.30.0 took 4 minutes 30 seconds. There's pretty much an exact 30 second increase in total time spent in infer_feature_types corresponding to is_pd_na

main

0.30.0

3. We need to perform is_pd_na but the trick is to only do it when the column is Unknown, that's the only case we care about. In python, the booleans are evaluated from left to right so we'll put the is_column_unknown check first since that one is near instant. Example:

4. With the new order, infer_feature_types now only takes 10 seconds longer compared to 0.30.0. This is coming from running convert_all_nan_unknown_to_double and checking if the columns are Unknown.

---

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 #2661 (0bff54c) into main (b0b657e) will not change coverage.
The diff coverage is 100.0%.

@@          Coverage Diff          @@
##            main   #2661   +/-   ##
=====================================
  Coverage   99.9%   99.9%           
=====================================
  Files        298     298           
  Lines      27305   27305           
=====================================
  Hits       27261   27261           
  Misses        44      44           

Impacted Files	Coverage Δ
evalml/utils/woodwork_utils.py	100.0% <100.0%> (ø)

---

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 b0b657e...0bff54c. Read the comment docs.

[chukarsten]

LGTM! Thanks for looking into this!

[angela97lin]

This is really cool and as always, your writeup is super clear. LGTM! 🚢 👏