Add event-taxonomy package with canonical schema, adapters, and CLI

Canonical NormalizedFinding schema with Severity enum (CRITICAL/HIGH/MEDIUM/LOW/INFO).
Per-project adapters for 9 tools with severity mapping for string labels, int 1-10,
float 0-1, Python Enum, and computed properties. CLI pipe interface and registry.

Nightshift-Task: event-taxonomy
Nightshift-Ref: https://github.com/marcus/nightshift

src/event_taxonomy/adapters/perf_regression.py

@@ -0,0 +1,16 @@
+from typing import Any
+
+from event_taxonomy.adapters._severity import map_int_severity
+from event_taxonomy.schema import NormalizedFinding
+
+
+def normalize(finding: dict[str, Any]) -> NormalizedFinding:
+    return NormalizedFinding(
+        tool="perf-regression-spotter",
+        category=finding.get("pattern", "performance"),
+        severity=map_int_severity(finding["severity"]),
+        message=finding["message"],
+        file=finding.get("file"),
+        line=finding.get("line"),
+        metadata={"snippet": finding["snippet"]} if finding.get("snippet") else {},
+    )