proplot: Slower speed of saving plotted Geographic data with cartopy

Description

I’m plotting large datasets using proplot (0.9.5.post78) and cartopy (0.20.0). The speed is much slower than the pyplot one.

Steps to reproduce

import time
import numpy as np
import proplot as pplt
import cartopy.crs as ccrs
import matplotlib.pyplot as plt

import warnings
warnings.filterwarnings('ignore')

offset = -40
lon = np.linspace(offset, 360 + offset - 1, 3000)
lat = np.linspace(-60, 60 + 1, 500)
state = np.random.RandomState(51423)
data = state.rand(len(lat), len(lon))

def proplot_test():
    start = time.time()
    fig, axs = pplt.subplots(proj='stere', figsize=(4,4))
    axs.coastlines()
    axs.pcolormesh(lon, lat, data, cmap='viridis')
    end = time.time()
    print('proplot plot cost: ', end - start)
    fig.savefig('test_proplot.jpg', dpi=300)
    save_end = time.time()
    print('proplot save cost: ', save_end - end)

def pyplot_test():
    start = time.time()
    fig, ax = plt.subplots(figsize=(4,4), subplot_kw=dict(projection=ccrs.Stereographic()))
    ax.coastlines()
    ax.pcolormesh(lon, lat, data, cmap='viridis', transform=ccrs.PlateCarree())
    end = time.time()
    print('pyplot plot cost: ', end - start)
    fig.savefig('test_pyplot.jpg', dpi=300)
    save_end = time.time()
    print('pyplot save cost: ', save_end - end)

proplot_test()
pyplot_test()

Expected behavior: [What you expected to happen]

The speed is similar with pyplot.

Actual behavior: [What actually happened]

proplot plot cost:  2.1709158420562744
proplot save cost:  5.2430949211120605

pyplot plot cost:  1.9175093173980713
pyplot save cost:  1.4209530353546143

In jupyterlab, the cost of plotting data by proplot is 7s while the pyplot one is 3.7s. I suppose this is the same problem of saving figures, because jupyter notebook needs to show the figure inline.