Phase 2B: Model architecture — DomainTransformerForCausalLM (NoPE, GPT-style), PLR embeddings, DCNv2 + JointFusion, 105 passing tests

Implements the full model architecture following Nubank nuFormer patterns:
- configuration.py: DomainTransformerConfig with presets (24M/85M/330M)
- modeling.py: GPT-style causal decoder with NoPE, SDPA attention, pre-norm, weight tying
- plr_embeddings.py: PeriodicLinearReLU numerical embeddings (Gorishniy et al. 2022)
- joint_fusion.py: DCNv2 + PLR + Transformer joint fusion (nuFormer-style)
- test_model.py: 33 tests covering config, model, PLR, DCNv2, joint fusion, integration
- All 105 tests passing (72 tokenizer + 33 model)

src/domain_tokenizer/models/__init__.py

@@ -0,0 +1,20 @@
+"""
+Model components for domainTokenizer.
+
+  - DomainTransformerConfig: HF-compatible configuration
+  - DomainTransformerForCausalLM: GPT-style causal decoder (NoPE)
+  - PeriodicLinearReLU: PLR numerical embeddings (Gorishniy et al. 2022)
+  - JointFusionModel: nuFormer-style Transformer + DCNv2 fusion
+"""
+
+from .configuration import DomainTransformerConfig
+from .modeling import (
+    DomainTransformerPreTrainedModel,
+    DomainTransformerModel,
+    DomainTransformerForCausalLM,
+    DomainTransformerAttention,
+    DomainTransformerMLP,
+    DomainTransformerBlock,
+)
+from .plr_embeddings import PeriodicLinearReLU
+from .joint_fusion import DCNv2CrossLayer, DCNv2, JointFusionModel