Description of proposed changes

In all Figure methods, the Session._preprocess method (or the Session._activate_figure method) is called before we call the actual plotting method. The Session._activate_figure method activates the figure of the current Figure object so that any subsequent plotting commands write to this figure. Calling the method is necessary when we have multiple Figure instances created and use them in a unfixed order. For example, it's necessary in the following case:

fig1 = pygmt.Figure()
fig2 = pygmt.Figure()
fig1.basemap(region=[0, 10, 0, 10], projection="X10c", frame=True)
fig2.basemap(region=[0, 10, 0, 10], projection="X10c", frame=True)
fig1.show()
fig2.show()

However, in cases where we only have one Figure instance or new Figure instance is created after we have finished the previous figure, calling Session._activate_figure is not needed. For example:

fig1 = pygmt.Figure()
fig1.basemap(region=[0, 10, 0, 10], projection="X10c", frame=True)
fig1.show()

fig2 = pygmt.Figure()
fig2.basemap(region=[0, 10, 0, 10], projection="X10c", frame=True)
fig2.show()

Thus, it would be good if we can only calling Session._activate_figure when necessary. This can be done by having a global PyGMT state variable that tracks the current activated figure. This PR serves as a proof of concept and it shows some performance improvements.

The PyGMT state variable was initially proposed in #2388.