the naming on the any_* fixtures in pandas/conftest.py is pretty confusing

• any_int_dtype (numpy only)
• any_int_nullable_dtype (EA only)
• any_numeric_dtype both numpy & EA
• any_real_dtype both numpy int & float
• sint_dtype signed numpy ints

for example

so we need dtypes that are numpy_only and EA only and numpy + EA both for ints & floats

would propose something like

any_int_numpy_dtype
any_int_ea_dtype
any_int_dtype -> both numpy + ea

same for float

then

any_numeric_nump_dtype
any_numeric_ea_dtype

for float + int

which means we rename
sint_dtype -> any_signed_int_numpy_dtype

I think this would require a PR to organize, a PR to create new names & fix everything.