Fixed deprecation warnings

[radarhere]

The Python 3.8 Travis job has started failing. A test for no unclosed file warning is failing on a deprecation warning that was recently added - python/cpython@0895793

This PR fixes the situation.

[hugovk]

Thanks! Good to see the 3.8-dev build has saved us surprises later!

[QuLogic]

Some of these look a bit suspicious to me. The types were changed from int to Py_ssize_t, but the PyArg_ParseTuple calls still specify i.

[radarhere]

@QuLogic isn't that the change of behaviour that the deprecation warning is describing though?

[QuLogic]

No, the #define only affects types that are # (e.g., y#); it does not affect i (which remains an int).

[QuLogic]

I've written a patch to change the PyArg_ParseTuple calls to n where necessary, but I see that only src/encode.c was really changed. Seemingly-equivalent variables in src/decode.c were not changed, e.g. GIF bits encode vs. decode. I'm fairly certain it will fix #3841, but I need to do more tests.

So would it be better to move to the n type specifier or revert some of these to int?