Change approach on how serialized files are handled

[fitoprincipe]

third attempt to fix #55

[12rambau]

My main problem with these changes is that you basically don't rely on the pytest-regression function anymore. This function even if it looks heavy is dealing with every edge cases of naming composition pytests offers and I would like to continue getting advantage of their support.

The only issue is that we cannot catch the real error that occurs when the file is not there or not corresponding (when the directly call pytest.raise) let me try to reach out to them first and see if they are open to modify this behaviour to a more specific error that we would be able to catch with the supress.

[fitoprincipe]

This is partially true; it's only avoiding pytest-regression when checking against the serialized object. This is the step when we need to make decisions, that's why I created the function that returns a boolean indicating only if the serialized object is the same as the saved one, avoiding all the extra condiments of pytest-regression. But still relies on it for the rest of the test.