fix(anthropic): drop redundant bt_safe_deep_copy passes in tracing (#583)

## Summary
- `SpanImpl.log_internal` already runs `bt_safe_deep_copy` on every
logged event (see `logger.py:4699`), so the eager copies in
`_normalize_anthropic_input`, `_get_input_from_kwargs`, and
`_redact_server_tool_output` were duplicating work on every span
open/log.
- Removed the three `bt_safe_deep_copy` calls and dropped the now-unused
import. The remaining walks (`_normalize_anthropic_data`,
`_process_input_attachments`, inlined redaction) already build fresh
dicts/lists and never mutate user data, so behavior is preserved.
- Mirrors what `google_genai/tracing.py` does today and matches the
`sdk-integrations` skill guidance: "Do not over-serialize. Braintrust
serializes at send/log time."

## Test plan
- [x] `pytest src/braintrust/integrations/anthropic/test_anthropic.py`
on anthropic==0.116.0 (matrix `latest`) — 48 passed
- [x] `pytest src/braintrust/integrations/anthropic/test_anthropic.py`
on anthropic==0.48.0 — 39 passed, 9 legitimately skipped for features
unavailable on that version
- No cassettes re-recorded, no new mocks/fakes added — used the existing
VCR-backed tests

🤖 Generated with [Claude Code](https://claude.com/claude-code)

<!-- sfk:created-approved-by -->
Created by abhijeet

<!-- sfk:slack-thread -->
[Slack
thread](https://starfolkai.slack.com/archives/C0AQDETAVT3/p1784236089663339?thread_ts=1784236089.663339&cid=C0AQDETAVT3)

Co-authored-by: Starfolk <noreply@starfolk.ai>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

Author: 3 people

py/src/braintrust/integrations/anthropic/tracing.py

@@ -3,7 +3,6 @@
 import warnings
 from typing import Any
 
-from braintrust.bt_json import bt_safe_deep_copy
 from braintrust.integrations.anthropic._utils import Wrapper, _try_to_dict, extract_anthropic_usage
 from braintrust.integrations.utils import _materialize_attachment
 from braintrust.logger import log_exc_info_to_span
@@ -904,7 +903,7 @@ def _normalize_anthropic_data(value: Any) -> Any:
 
 
 def _normalize_anthropic_input(value: Any) -> Any:
-    return _process_input_attachments(_normalize_anthropic_data(bt_safe_deep_copy(value)))
+    return _process_input_attachments(_normalize_anthropic_data(value))
 
 
 def _managed_agents_model_name(value: Any) -> str | None:
@@ -1280,10 +1279,9 @@ def _process_input_attachments(value):
 
 
 def _get_input_from_kwargs(kwargs):
-    msgs = bt_safe_deep_copy(list(kwargs.get("messages", [])))
-    msgs = _process_input_attachments(msgs)
+    msgs = _process_input_attachments(list(kwargs.get("messages", [])))
 
-    system = bt_safe_deep_copy(kwargs.get("system", None))
+    system = kwargs.get("system", None)
     if system:
         msgs.append({"role": "system", "content": _process_input_attachments(system)})
 
@@ -1343,19 +1341,14 @@ def _tool_span_input(call_item: dict[str, Any] | None) -> Any:
 
 
 def _redact_server_tool_output(value: Any) -> Any:
-    value = bt_safe_deep_copy(value)
-
-    def _redact(inner: Any) -> Any:
-        if isinstance(inner, list):
-            return [_redact(item) for item in inner]
-        if isinstance(inner, dict):
-            return {
-                key: ("<redacted>" if key in _SERVER_TOOL_SPAN_REDACTED_KEYS else _redact(item))
-                for key, item in inner.items()
-            }
-        return inner
-
-    return _redact(value)
+    if isinstance(value, list):
+        return [_redact_server_tool_output(item) for item in value]
+    if isinstance(value, dict):
+        return {
+            key: ("<redacted>" if key in _SERVER_TOOL_SPAN_REDACTED_KEYS else _redact_server_tool_output(item))
+            for key, item in value.items()
+        }
+    return value
 
 
 def _tool_span_output(result_item: dict[str, Any] | None) -> Any: