When importing mpl do not rely on submodules being loaded

That
```
import matplotlib as mpl
mpl.[submodule].[class]
```
works is an implementation detail because the top-level
namespace has imported these submodules. There's no
guarantee this will continue to work.

Also move the imports of the two relevant classes close
to their usage. This minimizes our top-level import to
the canonical `import matplotlib.pyplot as plt`.

Author: timhoffm

.flake8

@@ -47,6 +47,7 @@ per-file-ignores =
     lib/mpl_toolkits/axisartist/angle_helper.py: E221
 
     doc/conf.py: E402
+    galleries/users_explain/quick_start.py: E402
     galleries/users_explain/artists/paths.py: E402
     galleries/users_explain/artists/patheffects_guide.py: E402
     galleries/users_explain/artists/transforms_tutorial.py: E402, E501

galleries/users_explain/quick_start.py

@@ -17,7 +17,6 @@
 import numpy as np
 
 # sphinx_gallery_thumbnail_number = 3
-import matplotlib as mpl
 
 # %%
 #
@@ -446,13 +445,14 @@ def my_plotter(ax, data1, data2, param_dict):
 # well as floating point numbers.  These get special locators and formatters
 # as appropriate.  For dates:
 
+from matplotlib.dates import ConciseDateFormatter
+
 fig, ax = plt.subplots(figsize=(5, 2.7), layout='constrained')
 dates = np.arange(np.datetime64('2021-11-15'), np.datetime64('2021-12-25'),
                   np.timedelta64(1, 'h'))
 data = np.cumsum(np.random.randn(len(dates)))
 ax.plot(dates, data)
-cdf = mpl.dates.ConciseDateFormatter(ax.xaxis.get_major_locator())
-ax.xaxis.set_major_formatter(cdf)
+ax.xaxis.set_major_formatter(ConciseDateFormatter(ax.xaxis.get_major_locator()))
 
 # %%
 # For more information see the date examples
@@ -506,6 +506,8 @@ def my_plotter(ax, data1, data2, param_dict):
 # Often we want to have a third dimension in a plot represented by a colors in
 # a colormap. Matplotlib has a number of plot types that do this:
 
+from matplotlib.colors import LogNorm
+
 X, Y = np.meshgrid(np.linspace(-3, 3, 128), np.linspace(-3, 3, 128))
 Z = (1 - X/2 + X**5 + Y**3) * np.exp(-X**2 - Y**2)
 
@@ -518,8 +520,7 @@ def my_plotter(ax, data1, data2, param_dict):
 fig.colorbar(co, ax=axs[0, 1])
 axs[0, 1].set_title('contourf()')
 
-pc = axs[1, 0].imshow(Z**2 * 100, cmap='plasma',
-                          norm=mpl.colors.LogNorm(vmin=0.01, vmax=100))
+pc = axs[1, 0].imshow(Z**2 * 100, cmap='plasma', norm=LogNorm(vmin=0.01, vmax=100))
 fig.colorbar(pc, ax=axs[1, 0], extend='both')
 axs[1, 0].set_title('imshow() with LogNorm()')