fix(ac-0024): injective C identity via SHA-256 length-prefix digest + PostgreSQL migration/round-trip tests

Co-authored-by: genichin <4282227+genichin@users.noreply.github.com>

trex_core/application/reconciliation.py

@@ -25,6 +25,7 @@ published via AlarmEmitter with durable deduplication through ReconcileNotificat
 """
 from __future__ import annotations
 
+import hashlib
 import logging
 import math
 import re
@@ -105,12 +106,23 @@ def _normalize_legacy_cost_basis(value: object) -> tuple[float | None, str, str
 # ---------------------------------------------------------------------------
 
 def _c_identity_key(instance_key: str, symbol: str) -> str:
-    """Stable C-case notification identity key (PostgreSQL TEXT-safe, collision-resistant).
+    """Return an injective, PostgreSQL TEXT-safe key for a (instance_key, symbol) C-case pair.
 
-    Uses ASCII Unit Separator (\\x1f) which PostgreSQL TEXT accepts (only \\x00 is
-    forbidden) and which cannot appear in well-formed instance_key / symbol values.
+    Uses a length-prefixed SHA-256 hex digest so that *any* inputs — including
+    those containing separator characters — produce distinct keys with no
+    collision risk.  The prefix ``C:`` makes the key case-code-scoped and
+    human-scannable in the DB; the digest is 64 hex chars (256-bit).
+
+    Injectivity: the raw string fed to SHA-256 is ``"<N>:<instance_key>:<symbol>"``
+    where ``N = len(instance_key)``.  The leading ``N:`` uniquely delimits
+    ``instance_key`` from the rest, so two distinct ``(instance_key, symbol)``
+    pairs always produce different raw strings (and thus different digests).
+
+    Example: _c_identity_key("a\\x1fC", "b") != _c_identity_key("a", "C\\x1fb")
     """
-    return f"{instance_key}\x1fC\x1f{symbol}"
+    raw = f"{len(instance_key)}:{instance_key}:{symbol}"
+    digest = hashlib.sha256(raw.encode()).hexdigest()
+    return f"C:{digest}"
 
 
 def _build_alarm_fields(