llama : register the DT3 file type

Adds LLAMA_FTYPE_MOSTLY_DT3 at the end of the ftype enum, the loader
name/guess mappings, the quantization fallbacks (same as the other
ternary types), and the llama-quantize table entry. The table entry
warns that the in-tree quantizer is only the reference one: DT3 models
with the measured quality are produced by the external PTQTP pipeline.

tools/quantize/quantize.cpp

@@ -47,6 +47,7 @@ static const std::vector<quant_option> QUANT_OPTIONS = {
     { "IQ1_M",    LLAMA_FTYPE_MOSTLY_IQ1_M,    " 1.75 bpw quantization",            },
     { "TQ1_0",    LLAMA_FTYPE_MOSTLY_TQ1_0,    " 1.69 bpw ternarization",           },
     { "TQ2_0",    LLAMA_FTYPE_MOSTLY_TQ2_0,    " 2.06 bpw ternarization",           },
+    { "DT3",      LLAMA_FTYPE_MOSTLY_DT3,      " 3.5 bpw dual-ternary (REF quantizer only - use ternaria's PTQTP for quality)", },
     { "Q2_K",     LLAMA_FTYPE_MOSTLY_Q2_K,     " 2.96G, +3.5199 ppl @ Llama-3-8B",  },
     { "Q2_K_S",   LLAMA_FTYPE_MOSTLY_Q2_K_S,   " 2.96G, +3.1836 ppl @ Llama-3-8B",  },
     { "IQ3_XXS",  LLAMA_FTYPE_MOSTLY_IQ3_XXS,  " 3.06 bpw quantization",            },