Updated the drug categorization and improved manual curation

USER-FACING CHANGES

Improved drug categorization

drugs_to_drug_classes() also uses the 'has_part' ARO relationship now to get drug classes for antibiotic mixtures. In case of antibiotic mixtures, the drug classes of the drugs associated with 'has_part' are returned rather than 'antibiotic mixture' (ARO:3000707).
'antibiotic mixture' will not be reported as a drug class, rather the individual antibiotic classes making up the antibiotic mixture will be reported.

manual curation (argannot): (Tet)tetH:EF460464:6286-7839:1554 was incorrectly annotated as ARO:3004797 which is a beta-lactamase due to a loose RGI hit. This was manually curated to ARO:3000175.
Improved curation:
- resfinder_curation: grdA_1_QJX10702 -> 3007380 & EstDL136_1_JN242251 -> 3000557
- megares_curation: MEG_2865|Drugs|Phenicol|Chloramphenicol_hydrolase|ESTD -> 3000557

confers_resistance_to() now gets drugs information even if it is encoded at a higher level in the ARO. For example, OXA-19 previously only returned cephalosporin and penam, but now will also return oxacillin (from AMR gene family).
drugs_to_drug_classes() now correctly only returns the immediate child of 'antibiotic molecule' as the drug class (this was previously not the case for certain corner cases).
inconsistent ARO versions deeparg, megares, resfinderfg & sarg curation: ARO:3004445 -> ARO:3005440, this was due to a change in the ARO and the ARO number for the RSA2 gene changing, but the version of ARO bundled with argNorm was out of sync.

AROs were previously handled as integers in the get_aro_mapping_table() function and this posed challenges when ARO numbers such as 'ARO:0010004' (leading zeros leading to issues). To fix this, AROs are now treated as strings so leading zeros can be maintained.