Normal.free_symbols vs Indexed.free_symbols

[smichr]

This is related to #14899. Compare Normal and Indexed response to free symbols queries:

>>> from sympy.stats import Normal
>>> X = Normal('X', 2, 4)
>>> [(i,type(i)) for i in X.free_symbols]
[(X, <class 'sympy.stats.rv.RandomSymbol'>)]
>>> [(i,type(i)) for i in X.expr_free_symbols]
[(X, <class 'sympy.core.symbol.Symbol'>)]

vs

>>> Fx = IndexedBase("F")[x]
>>> [(i,type(i)) for i in Fx.expr_free_symbols]
[(F[x], <class 'sympy.tensor.indexed.Indexed'>)]
>>> [(i,type(i)) for i in Fx.free_symbols]
[(F[x], <class 'sympy.tensor.indexed.Indexed'>), (x, <class 'sympy.core.symbol.Symbol'>), (F, <class 'sympy.core.symbol.Symbol'>)]

It seems like Normal should give this:

>> [(i,type(i)) for i in X.expr_free_symbols]
[(X, <class 'sympy.stats.rv.RandomSymbol'>)]
>> [(i,type(i)) for i in X.free_symbols]
[(X, <class 'sympy.stats.rv.RandomSymbol'>), (X, <class 'sympy.core.symbol.Symbol'>)]