BUG: yt may leak matplotlib configuration globally at runtime

[neutrinoceros]

Bug report

Bug summary

Matplotlib contour styles for monochrome contour sets can be changed globally by yt at runtime (this is triggered by importing from yt.visualization._mpl_imports, which isn't implied by import yt).
The problematic line is very obviously:

yt/yt/visualization/_mpl_imports.py

Line 10 in e525381

matplotlib.rc("contour", negative_linestyle="solid")

(in fact I spotted the problem in the code first, and then worked on a script to show its undesired effect)

Code for reproduction

# adapted from https://matplotlib.org/stable/gallery/images_contours_and_fields/contour_label_demo.html#sphx-glr-gallery-images-contours-and-fields-contour-label-demo-py
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
import yt

delta = 0.025
x = np.arange(-3.0, 3.0, delta)
y = np.arange(-2.0, 2.0, delta)
X, Y = np.meshgrid(x, y)
Z1 = np.exp(-X**2 - Y**2)
Z2 = np.exp(-(X - 1)**2 - (Y - 1)**2)
Z = (Z1 - Z2) * 2
zm = Z.mean()
Zp = Z - zm

def make_figure(filename):
    # Basic contour plot
    fig, ax = plt.subplots()
    CS = ax.contour(X, Y, Zp, linestyles=None, colors="black")
    #assert CS.negative_linestyles == "dashed"

    print(f"saving to {filename}")
    fig.savefig(filename)

# draw the figure
make_figure("/tmp/pre.png")

# do stuff with yt
ds = yt.load_sample("IsolatedGalaxy")
yt.SlicePlot(ds, "z", "density").save("/tmp/")

# try the pure-mpl figure again
make_figure("/tmp/post.png")

Actual outcome

Expected outcome

pure matplotlib graphs should not be affected by yt at runtime.