docs: clarify strict-mode refusal message and to_bytes inlining docs

Address PR review feedback:

- codec.rs: rewrite strict-refusal error to present the two real
  remediations (sender re-encode by-name + receiver register; or
  receiver enables inlining, accepting cloudpickle risk) instead of
  bundling registration with both-side inlining.
- expr.py: qualify to_bytes docstring so Python UDF self-contained
  behavior is conditional on with_python_udf_inlining being enabled.

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

Author: timsaucer

crates/core/src/codec.rs

@@ -413,10 +413,12 @@ fn refuse_inline_payload(kind: &str, name: &str) -> datafusion::error::DataFusio
     // error would mis-route it into "fix your SQL" buckets.
     datafusion::error::DataFusionError::Execution(format!(
         "Refusing to deserialize inline Python {kind} '{name}': Python UDF \
-         inlining is disabled on this session. Ask the sender to re-encode \
-         with inlining disabled (so the UDF travels by name), or register \
-         '{name}' on this receiver's session and enable inlining on both \
-         sides — receivers cannot re-encode bytes they did not produce."
+         inlining is disabled on this session. Two remediations: \
+         (1) ask the sender to re-encode with inlining disabled so '{name}' \
+         travels by name, and register '{name}' on this receiver; or \
+         (2) enable inlining on this receiver (accepts the cloudpickle \
+         execution risk on inbound payloads). Receivers cannot re-encode \
+         bytes they did not produce."
     ))
 }
 

python/datafusion/expr.py

@@ -452,10 +452,13 @@ def to_bytes(self, ctx: SessionContext | None = None) -> bytes:
         When ``ctx`` is ``None``, the default codec is used (Python UDF
         inlining on, no user-installed extension codec).
 
-        Built-in functions and Python UDFs (scalar, aggregate, window)
-        travel inside the returned bytes; the worker does not need to
-        pre-register them. UDFs imported via the FFI capsule protocol
-        travel by name only and must be registered on the worker.
+        Built-in functions travel inside the returned bytes. Python UDFs
+        (scalar, aggregate, window) also inline by default, so the worker
+        does not need to pre-register them; when the encoding session has
+        :meth:`SessionContext.with_python_udf_inlining` set to ``False``,
+        Python UDFs travel by name only and must be registered on the
+        worker. UDFs imported via the FFI capsule protocol always travel
+        by name only and must be registered on the worker.
 
         .. warning:: Security
             Bytes returned here may embed a cloudpickled Python