lr_finder plotting method automated unit tests

[wyli]

Describe the bug

MONAI/tests/test_lr_finder.py

Line 76 in e0feb53

lr_finder.plot(0, 0) # to inspect the loss-learning rate graph

this plotting test may block the main testing runner and need manual confirmation

To Reproduce
./runtests.sh --quick --unittest

Stopping early, the loss has diverged                                                                                                                                                                                                             
Computing optimal learning rate:  80%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏                                  | 4/5 [00:06<00:01,  1.72s/it]
Resetting model and optimizer
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
Coverage.py warning: No data was collected. (no-data-collected)
0.00031622776601683794
??^B^C^C
^C
^C

^C
^C^C
^C

^CFinished test: test_lr_finder (tests.test_lr_finder.TestLRFinder) (2.63e+03s)

[rijobro]

Do you think something like this might solve the problem? https://stackoverflow.com/a/40395799/12411867

We can pass ax to lr_finder.plot(0, 0), so we can create the plot prior to passing it to lr_finder.

[wyli]

sounds good, not sure if it's feasible to follow the idea in the other test

MONAI/tests/test_threadcontainer.py

Line 97 in e0feb53

fig = con.plot_status(logger)