test(python): Use assert_frame_equal instead of assert df.frame_equal(...)

[stinodego]

Relates to #6364

assert_frame_equal is more powerful as it checks dtypes as well, and has better error messages.

I found a few bugs (these are now fixed in different PRs), as well as found a few tests that weren't asserting the equality (simply df.frame_equal(expected) without the assert 😄 )

Random interesting tidbit

Different default dtypes for numpy arrays between Linux/Windows

# On Windows
>>> np.array([1, 2, 3]).dtype
dtype('int32')

# On Linux
>>> np.array([1, 2, 3]).dtype
dtype('int64')

This broke some tests, which I fixed by explicitly specifying the numpy dtype as np.int64.

[stinodego]

@ritchie46 Can this be merged? It's pretty big so very sensitive to merge conflicts.

[ritchie46]

Yes! And thanks for the cleanup. It did expose some bugs.