feat(V2): /mcp-cross/tools/get_historical with cross-exchange consensus

Add a unified historical endpoint that fans out to every exchange
supporting the requested (asset_class, symbol) pair, then merges the
results into a single consensus candle series with per-bar divergence
metrics:
  - candles[i].close = median across sources
  - candles[i].sources = count of contributing exchanges
  - candles[i].div_pct = (max-min)/median for that bar's close

Crypto routes BTC/ETH/SOL across Bybit + Hyperliquid + Deribit; equities
route to Alpaca for now (IBKR omitted from MVP because its bars endpoint
takes a relative period instead of start/end). Partial failures return a
warning envelope (failed_sources) instead of failing the whole request;
all sources failing → HTTP 502.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

tests/unit/exchanges/cross/test_consensus.py

@@ -0,0 +1,90 @@
+from __future__ import annotations
+
+from cerbero_mcp.exchanges.cross.consensus import merge_candles
+
+
+def _c(ts, o, h, l, c, v):
+    return {"timestamp": ts, "open": o, "high": h, "low": l, "close": c, "volume": v}
+
+
+def test_empty_input():
+    assert merge_candles({}) == []
+
+
+def test_single_source_passthrough():
+    out = merge_candles({"bybit": [_c(1, 100, 110, 90, 105, 5)]})
+    assert len(out) == 1
+    assert out[0]["timestamp"] == 1
+    assert out[0]["close"] == 105
+    assert out[0]["sources"] == 1
+    assert out[0]["div_pct"] == 0.0
+
+
+def test_three_sources_identical_no_divergence():
+    src = {
+        "bybit":       [_c(1, 100, 110, 90, 105, 5)],
+        "hyperliquid": [_c(1, 100, 110, 90, 105, 3)],
+        "deribit":     [_c(1, 100, 110, 90, 105, 7)],
+    }
+    out = merge_candles(src)
+    assert len(out) == 1
+    assert out[0]["close"] == 105.0
+    assert out[0]["sources"] == 3
+    assert out[0]["div_pct"] == 0.0
+    # volume is mean across sources
+    assert abs(out[0]["volume"] - 5.0) < 1e-9
+
+
+def test_three_sources_divergent_close():
+    src = {
+        "bybit":       [_c(1, 100, 110, 90, 100, 1)],
+        "hyperliquid": [_c(1, 100, 110, 90, 110, 1)],
+        "deribit":     [_c(1, 100, 110, 90, 105, 1)],
+    }
+    out = merge_candles(src)
+    # median of [100, 110, 105] = 105
+    assert out[0]["close"] == 105.0
+    # div_pct = (110 - 100) / 105 ≈ 0.0952
+    assert abs(out[0]["div_pct"] - 10 / 105) < 1e-6
+    assert out[0]["sources"] == 3
+
+
+def test_misaligned_timestamps():
+    src = {
+        "bybit":       [_c(1, 100, 110, 90, 105, 1), _c(2, 100, 110, 90, 105, 1)],
+        "hyperliquid": [_c(2, 100, 110, 90, 105, 1), _c(3, 100, 110, 90, 105, 1)],
+    }
+    out = merge_candles(src)
+    timestamps = [c["timestamp"] for c in out]
+    sources_by_ts = {c["timestamp"]: c["sources"] for c in out}
+    assert timestamps == [1, 2, 3]
+    assert sources_by_ts == {1: 1, 2: 2, 3: 1}
+
+
+def test_two_sources_even_median():
+    src = {
+        "bybit":       [_c(1, 100, 110, 90, 100, 1)],
+        "hyperliquid": [_c(1, 100, 110, 90, 110, 1)],
+    }
+    out = merge_candles(src)
+    # even median = mean of two = 105
+    assert out[0]["close"] == 105.0
+
+
+def test_empty_source_ignored():
+    src = {
+        "bybit":       [_c(1, 100, 110, 90, 105, 1)],
+        "hyperliquid": [],
+    }
+    out = merge_candles(src)
+    assert len(out) == 1
+    assert out[0]["sources"] == 1
+
+
+def test_output_sorted_by_timestamp():
+    src = {
+        "bybit": [_c(3, 100, 110, 90, 105, 1), _c(1, 100, 110, 90, 105, 1),
+                  _c(2, 100, 110, 90, 105, 1)],
+    }
+    out = merge_candles(src)
+    assert [c["timestamp"] for c in out] == [1, 2, 3]