How to align a legend relative to a GridSpec cell? - matplotlib

Question

I am creating a figure like this:
fig = plt.figure(figsize = (7, 8))
outer_grid = gridspec.GridSpec(2, 1, height_ratios = [2, 1])
inner_grid1 = gridspec.GridSpecFromSubplotSpec(4, 3, subplot_spec=outer_grid[0])
inner_grid2 = gridspec.GridSpecFromSubplotSpec(2, 3, subplot_spec=outer_grid[1])
Now I would like to have one legend for all plots in inner_grid1 and a separate legend for all plots in inner_grid2. And I would like those legends to be placed nicely, even though they are higher than a single plot, and cannot have more than one column to not make the figure too wide.
Here is an example where I tried to align the legends with trial and error with method 2 below, however this took ages to make.
So I see three options to achieve this, none of which work:
Place the legend as part of an Axes object, but manually move it outside of the actual plot using axes.legend([...], bbox_to_anchor=(x, y)). This does not work when the legend is higher as a single plot, because it rescales the plots to fit the legend into its grid cell.
Place the legend globally on the Figure object. This works, but makes the correct placement really hard. I cannot use loc = "center right", since it centers it for the full figure instead of just the inner_grid1 or inner_grid2 plots.
Place the legend locally on the GridSpecFromSubplotSpec object. This would be perfect. However there is no method to create a legend on a GridSpecFromSubplotSpec or related classes, and the pyplot.legend method misses parameters to restrict the loc to parts of a grid.
Is there a way to place a legend as described?
As requested, a small code example generating something similar as desired.
This example uses method 2:
#!/usr/bin/env python3
import pandas as pd, seaborn as sns
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
GENOMES = ["spneumoniae", "ecoliK12", "scerevisiae", "celegans", "bmori", "hg38"]
fig = plt.figure(figsize = (7, 8))
outer_grid = gridspec.GridSpec(2, 1, height_ratios = [2, 1])
inner_grid1 = gridspec.GridSpecFromSubplotSpec(4, 3, subplot_spec=outer_grid[0])
inner_grid2 = gridspec.GridSpecFromSubplotSpec(2, 3, subplot_spec=outer_grid[1])
# plots are in sets of six, 2 rows by 3 columns each
for index, genome in enumerate(GENOMES):
data = pd.DataFrame({"x": [0, 1, 2, 3, 0, 1, 2, 3], "y": [1, 0, 3, 2, 1, 0, 3, 2], "hue": ["a", "a", "a", "a", "b", "b", "b", "b"]})
# first set of six
ax1 = plt.Subplot(fig, inner_grid1[index])
ax1 = sns.lineplot(data = data, x = "x", y = "y", hue = "hue", ax = ax1)
ax1.set_xlabel("")
ax1.set_ylabel("")
if index == 2:
ax1.legend()
handles, labels = ax1.get_legend_handles_labels()
fig.legend(handles, labels, loc = "center left", title = "", bbox_to_anchor=(0.9, 2/3 - 0.03))
ax1.legend([], [], loc = "lower center", title = f"{genome}")
fig.add_subplot(ax1)
# second set of six
ax2 = plt.Subplot(fig, inner_grid1[index + 6])
ax2 = sns.lineplot(data = data, x = "x", y = "y", hue = "hue", ax = ax2)
ax2.set_xlabel("")
ax2.set_ylabel("")
ax2.legend([], [], loc = "upper center", title = f"{genome}")
fig.add_subplot(ax2)
#third set of six
ax3 = plt.Subplot(fig, inner_grid2[index])
ax3 = sns.lineplot(data = data, x = "x", y = "y", hue = "hue", ax = ax3)
ax3.set_xlabel("")
ax3.set_ylabel("")
if index == 2:
ax3.legend(["#unitigs", "avg. unitig len."])
handles, labels = ax3.get_legend_handles_labels()
fig.legend(handles, labels, loc = "center left", title = "", bbox_to_anchor=(0.9, 1/6 + 0.05))
ax3.legend([], [], loc = "upper center", title = f"{genome}")
fig.add_subplot(ax3)
plt.savefig("stackoverflow_test.pdf", bbox_inches="tight")