use 'is not None' for max_nonstreaming_tokens check in timeout calculation

[rmotgi1227]

Problem

_calculate_nonstreaming_timeout uses a truthiness guard on max_nonstreaming_tokens:

# _base_client.py line 736
if expected_time > default_time or (max_nonstreaming_tokens and max_tokens > max_nonstreaming_tokens):

The parameter is typed int | None. Checking truthiness silently skips the guard when the value is 0 — a value that would mean "streaming is always required regardless of token count." The condition evaluates to False and no ValueError is raised.

Fix

if expected_time > default_time or (max_nonstreaming_tokens is not None and max_tokens > max_nonstreaming_tokens):

is not None matches the type annotation, makes the intent explicit, and correctly handles a hypothetical max_nonstreaming_tokens=0 entry in MODEL_NONSTREAMING_TOKENS.