PyTokenizer_FindEncoding() always failed because it set the tokenizer state

PyTokenizer_FindEncoding() always failed because it set the tokenizer state
with only a file pointer when it called fp_setreadl() which expected a file
path. Changed fp_setreadl() to use either a file path or file descriptor
(derived from the file pointer) to fix the issue.

Closes issue 3594.
Reviewed by Antoine Pitrou and Benjamin Peterson.