json_normalize - incorrect removing separator from beginning of key

[iyonchev]

print(pd.__version__)
jsn={'_id':{'a1':10,"l2":{"l3":0}}, "gg":4}
display(pd.json_normalize(jsn,sep='_'))

Actual result:

1.5.2

gg	id_a1	id_l2_l3
4	10	0

Expected result:
the "_" prefix should not be removed from the beginning of the column names.

1.5.2

gg	_id_a1	_id_l2_l3
4	10	0

This issue was introduced with the following change:
http://github.com/pandas-dev/pandas/pull/40035/files

[iyonchev]

display(pd.json_normalize(jsn,sep='_', max_level=5))
passing max_level value could be used to prevent the execution of the new methods(from the change above) and execute the main one...

[jbrockmendel]

Cc @WillAyd