Fixes cover the following:
- --pheno/--covar loading of categorical data when the input file doesn't include all samples in the main fileset
- --glm logistic-regression handling of mixed no-dosages-missing and some-dosages-missing data (results could be nondeterministically slightly off)
- PCA projection workflow:
- "--pca allele-wts" scaling of multiallelic variant weights
- --score center/variance-standardize missing-dosage imputation
- --score variance-standardize tolerance of monomorphic variants
- Several merge CM-column and conflict-handling bugs
- "--make-pgen multiallelics=-" phase handling
- "--bgen ref-first" phase-handling (everything was swapped)
- "--export bcf vcf-dosage=HDS-force" on unphased data
- "--bcf dosage=HDS" on one-value (haploid) HDS fields
- "--export vcf vcf-dosage=DS-force" on multiallelic chrX variants
- "--export vcf vcf-dosage=DS-{force,only}" on multiallelic variants
- Import of old (v1.1) BGEN files containing both dosages and alternate contigs
- Handling of extremely long --q-score-range names
- chrY allele counts/frequencies when unknown-sex samples are present
- Swapped HET1_HOM2/HET2_HOM1 in --make-king-table output
- --adjust-file on input without a PROVISIONAL_REF? column
- Detection/handling of corrupted .pgen files
- Error message on VCF/BCF files with too many (more than (2^31 - 3)) variants for PLINK 2