P6: gtc-to-vcf (VCF/BCF + indexes), harmonisation, exports (plink, bedgraph, finalreport)

- formats/fasta: plain/BGZF FASTA, .fai/.gzi built identical to htslib, index check, build detection,
  contig aliases, mmap genome cache (IN-060..064)
- formats/bgzf, formats/bcf: pure-Python BGZF, tabix .tbi, BCF 2.2 and CSI (OUT-010, OUT-038, OUT-040)
- harmonize/sites: plus-strand mapping with audit rows, indel resolution from SourceSeq (port of GTCtoVCF,
  Apache-2.0) and left normalisation (OUT-014..021, OUT-030, OUT-042)
- harmonize/combine: OUT-031 genotype combination (ploidy-aware, Infinium I/II signal classes, OUT-037)
- pipeline/vcf: `simurg gtc-to-vcf` (per-sample VCF.gz+tbi, --merge cohort BCF+csi, audit parquet)
- pipeline/export: `simurg export bedgraph|finalreport|plink|locus-summary` (OUT-043, OUT-050..052, OUT-071)
- validation: 36 GSA samples vs gtc2vcf: 23.66 M identical calls, 0 unexplained differences; PLINK identical
  to plink2 on 637,397 variants; bcftools validates VCF/BCF and our indexes answer region queries correctly
- fetch_data.sh p6 (Ensembl GRCh38 with checksum), make_oracle_vcf.sh, validate_p6_vcf.py; K-022/K-023

src/simurg/core/errors.py

@@ -89,6 +89,28 @@ CATALOGUE: dict[str, MessageDef] = dict(
         _m("E_EGT_LOAD", "error", "Cluster (EGT) file cannot be loaded", "IN-044"),
         _m("W_EGT_NAME", "warning", "Cluster file name is not portable", "IN-045"),
         _m("W_MANIFEST_NO_REFSTRAND", "warning", "Manifest lacks RefStrand annotation", "IN-040"),
+        # --- reference genome and VCF export (IN-060 .. IN-065, OUT-010 .. OUT-043) --------------------------
+        _m("E_FASTA_FORMAT", "error", "FASTA file is malformed", "IN-060"),
+        _m("E_FAI_MISMATCH", "error", "FASTA index does not match the FASTA", "IN-060"),
+        _m("E_BUILD_MISMATCH", "error", "Manifest genome build differs from the FASTA build", "IN-063"),
+        _m("W_SKIP_UNMAPPED", "warning", "Locus without a genomic position was skipped", "OUT-019"),
+        _m("W_SKIP_INTENSITY_ONLY", "warning", "Intensity-only locus was skipped", "OUT-019"),
+        _m("W_SKIP_INDEL_CONTEXT", "warning", "Indel sequence cannot be determined (no CSV SourceSeq)", "OUT-019"),
+        _m("W_SKIP_INDEL_REF", "warning", "Indel reference allele cannot be determined from the FASTA", "OUT-019"),
+        _m("W_INDEL_PARTIAL_MATCH", "warning", "Indel source sequence matches the genome only partially", "OUT-019"),
+        _m("W_SKIP_EXCLUDED", "warning", "Locus excluded by --exclude-loci/--include-loci", "OUT-019"),
+        _m("W_SKIP_CONTIG", "warning", "Manifest chromosome absent from the FASTA", "OUT-019"),
+        _m("W_SKIP_NO_STRAND", "warning", "Locus has neither RefStrand nor a sequence to align", "OUT-042"),
+        _m("W_SKIP_AMBIGUOUS", "warning", "Ambiguous-strand locus dropped by --drop-ambiguous", "OUT-042"),
+        _m("W_REF_NOT_QUERIED", "warning", "The reference allele is not among the probe's alleles", "OUT-015"),
+        _m("W_MERGE_INCONSISTENT", "warning", "Probes at a site disagree; genotype set missing", "OUT-031"),
+        _m("W_MERGE_AMBIGUOUS", "warning", "Probes at a site allow several genotypes; set missing", "OUT-031"),
+        _m("W_MERGE_ALL_NC", "warning", "Every probe at a merged site is a no-call", "OUT-031"),
+        _m("W_INF1_NC_CONFLICT", "warning", "Infinium I no-call conflicts with other assays", "OUT-037"),
+        _m("W_NO_BGZIP", "warning", "Uncompressed VCF written; downstream commands need bgzip + index", "OUT-010"),
+        _m("W_UNKNOWN_LOCI", "warning", "Names in a locus list are not in the manifest", "OUT-036"),
+        _m("E_AUX_INVALID", "error", "Auxiliary-loci record is invalid", "OUT-035"),
+        _m("W_LARGE_OUTPUT", "warning", "Output will be very large", "EDGE-036"),
         # --- QC (QC-001 .. QC-064) -----------------------------------------------------------------
         _m("W_CTRL_MISSING", "warning", "Controls needed by a metric are absent; metric not evaluable", "QC-007"),
         _m("W_CTRL_ZERO_DIVISOR", "warning", "Control metric has a zero denominator (NaN)", "QC-008"),