Fix handle memory leaks
Currently the alpm handle is released while db's still implicitly reference to it making a simple program which opens a handle and get_syncdbs(). Somewhere along Python makes the handle go out of scope which releases the alpm_handle.
The db's then use it and pyalpm segfaults. What should happen is that the DB get's an ALPM handle object and incref's it and when the DB goes out of scope DECREF's it. This should resolve the alpm_release being called too soon, since DB's still reference to the handle object.
Test all features of https://github.com/actionless/pikaur before releasing a memory leak fixed version. Same goes for prometheus-arch-exporter, namcap, nvchecker and arch's security-tracker