Make HistoryEncoder a pluggable registry, like Router/Objective (gitea #35)

Stage2Autoregressive.init_history_cache and .history_step both
isinstance-checked self.history_encoder against AttentionHistory to decide
whether to use its real incremental-cache methods or a no-op fallback, so a
third history type couldn't be added without editing Stage2Autoregressive
itself. The two-value "markov"/"attention" enum was also independently
hardcoded in three places (Stage2Autoregressive's own validation,
config.py's validate_config, and AutoregressiveConfig.from_dict's default).

Mirrors the Router (giant/model/routers.py) and Objective
(giant/model/objectives.py, gitea #32) pattern: HistoryEncoder now declares
working O(1) init_cache/step defaults (init_cache -> None, step -> one
forward() call), so every registered history type satisfies the incremental
interface without opting in; AttentionHistory overrides both with its real
KV-cache versions since its forward() needs the full prefix. Added
HISTORY_REGISTRY/register_history/build_history, registered "markov" and
"attention", and deleted both isinstance checks in models.py.

Per user decision during planning, config.py's validate_config now imports
HISTORY_REGISTRY and checks membership dynamically instead of keeping its own
hardcoded tuple, making the registry the single source of truth end to end
(verified no import cycle: config.py had no prior dependency on giant.model,
and giant.model.history has none on giant.config).

No config-schema change and no checkpoint impact — this is a pure
internal-interface refactor.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

tests/test_network.py

@@ -5,14 +5,17 @@ import torch
 from giant import config as gconfig
 from giant.constants import CONT_SLOT_DIM, COND_DIM, PARTICLE_PHYS_DIM, SEC_SLOT_DIM
 from giant.model.network import (
+    HISTORY_REGISTRY,
     AttentionHistory,
     ConditionEncoder,
+    HistoryEncoder,
     MarkovHistory,
     SinusoidalEmbedding,
     Stage1Model,
     Stage2Autoregressive,
     Stage2OneShot,
     build_critics,
+    build_history,
     build_models,
     cat_col_layout,
     stage2_trunk_sec_dim,
@@ -407,6 +410,48 @@ def test_attention_history_step_matches_forward():
     assert torch.allclose(stepped, expected, atol=1e-5)
 
 
+# --- HISTORY_REGISTRY / build_history (gitea #35) ----------------------------
+
+
+def test_history_registry_has_exactly_the_two_known_histories():
+    assert set(HISTORY_REGISTRY) == {"markov", "attention"}
+
+
+def test_build_history_returns_correct_concrete_type():
+    assert isinstance(build_history("markov", 4, 6), MarkovHistory)
+    assert isinstance(build_history("attention", 4, 8), AttentionHistory)
+
+
+def test_build_history_unknown_name_raises():
+    with pytest.raises(ValueError):
+        build_history("bogus", 4, 6)
+
+
+def test_build_history_filters_kwargs_by_signature():
+    """Attention-only kwargs (n_heads/n_layers) must be silently dropped when
+    building a MarkovHistory, matching build_router's documented behavior for
+    per-type hyperparameters coexisting in one config."""
+    hist = build_history("markov", 4, 6, n_heads=2, n_layers=1)
+    assert isinstance(hist, MarkovHistory)
+
+
+def test_history_encoder_base_default_init_cache_and_step():
+    """A HistoryEncoder subclass implementing only forward() must still get
+    working O(1) init_cache/step defaults from the base class."""
+
+    class _StubHistory(HistoryEncoder):
+        def forward(self, feat, has_prev):
+            return feat * 2
+
+    hist = _StubHistory()
+    assert hist.init_cache() is None
+    feat = torch.randn(2, 1, 4)
+    has_prev = torch.ones(2, 1, dtype=torch.bool)
+    out, cache = hist.step(feat, has_prev, "unused-cache")
+    assert torch.equal(out, hist.forward(feat, has_prev))
+    assert cache == "unused-cache"
+
+
 # --- Stage2Autoregressive (v0.3.0 step 5) -----------------------------------