Memory leak with pyplot

When I create matplotlib.pyplot plots, the memory keeps increasing and is not released by the garbage collector. The following is the case I'm interested in (with subplots) and bellow I include a more basic MWE.

Example with subplots

```julia
using PyCall
plt = pyimport("matplotlib.pyplot")

function test_pyplot()
for i in 1:10000
@info i
z = rand(4, 50)
ẑ = rand(4, 50)
fig, axs = plot_pyplot(z, ẑ)
end
end

function plot_pyplot(z, ẑ)
# plt = pyimport("matplotlib.pyplot")
N = Int(size(z, 1)/2)
fig, axs = plt.subplots(2, N)
for i in 1:N
axs[1,i].plot(z[i,:], label = "true")
axs[1,i].plot(ẑ[i,:], label = "model")

axs[2,i].plot(z[i+N,:], label = "true")
axs[2,i].plot(ẑ[i+N,:], label = "predicted")

axs[2,i].set_xlabel("time steps")
axs[i].grid(true)
axs[i+N].grid(true)

axs[i+N].label_outer()
axs[i].label_outer()
end
# plt.close()
return fig, axs
end

test_pyplot()
```

I tried defining `plt` and closing it inside the `plot_pyplot` function (commented lines) but it doesn't solve the memory leak.

Here you have a simpler MWE:
```julia
using PyCall
plt = pyimport("matplotlib.pyplot")

plot_pyplot(z) = plt.plot(z)

function test_pyplot()
for i in 1:1000000
@info i
z = rand(50)
fig = plot_pyplot(z)
end
end
```

Version info and status

```julia
julia> versioninfo()
Julia Version 1.8.2
Commit 36034abf260 (2022-09-29 15:21 UTC)
Platform Info:
OS: macOS (arm64-apple-darwin21.3.0)
CPU: 8 × Apple M1 Pro
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-13.0.1 (ORCJIT, apple-m1)
Threads: 1 on 6 virtual cores
```

```julia
(memory_leak) pkg> st
Status `~/Documents/issues/memory_leak/Project.toml`
[438e738f] PyCall v1.94.1
```

I need to use pyplot directly from PyCall.jl and not through PyPlot.jl because I need to pass the fig object to Wandb.jl. Actually, maybe I could get that object from PyPlot.jl, I didn't try that yet. But since I found the bug here, I think it can be helpful to report it anyway.

Beginne mit dem bereitgestellten Julia MWE und den PyCall-Aufrufen zu pyplot und vergleiche die Speichernutzung mit und ohne Erstellung von Plots und Garbage Collection. Verfolge die zurückgegebenen fig- und plot-Objekte durch PyCall; abgeschlossen ist die Aufgabe, wenn die beibehaltenen Referenzen identifiziert und behoben sind, sodass wiederholtes Plotten nicht mehr zu unbegrenztem Speicherwachstum führt.

Vom Indexierungsmodell aus dem Issue-Text verfasst.