Changes were made in the scripts to resolve errors in the data cleani…

…ng and formatting of the PharmGKB CSV+tMCF pairs. It also breaks out the phenotypes to distinguish that one is a MeSHQualifier and seven are MeSHSupplementaryConceptRecords, unlike the rest of the phenotypes which are MeSHDescriptors. Therefore these were separated into 3 CSV+tMCF pairs. In particular, the links to the enums and between entity types were fixed. This was done by initializing all nodes referenced and then pointing to them within the tMCF. Because of this any existence missing errors in the json reports can be ignored. The changes to the scripts, tMCF files, and documentation (README.md) for this import are part of GitHub PR 1056 datacommonsorg/data#1056

Schema Changes:
- Add CPICLevelEnum, DosageGuidelineSourceCpicNoRecommendation, DrugTypeEnum, PGxLevelEnum, PharmacogeneticAssociationEnum.
- Add properties for clinicalAnnotationCount, clinicalAnnotationCountLevel1_2, clinicalGuidelineAnnotationCount, dosageGuideline, drugHasPrescribingInfo, drugLabelAnnotationCount, drugType, fdaTopPharmacogeneticLevel, geneticVariantAnnotationCount, hasCpicDosingGuideline, hasGenomicCoordinates, hasGeneticVariantAnnotation, hasPrescribingInfo, medicalDictionaryForRegulatoryActivitiesId, metabolicPathwayCount, pharmageneticAssociation, topClinicalAnnotationLevel, topCpicLevel, topPharmacogeneticLevel, veryImportantPharmacogeneCount.
- Remove properties for fdaTopPGxLevel, mintID, nationalClinicalTrialNumber, nationalDrugCode, nationalDrugFileReferenceTerminologyCode, neuroMabID, patentID,pharmGkbClinicalAnnotationCount,pharmGkbPathwayCount, pkgbTags.

PiperOrigin-RevId: 653311391