fix(agent): coerce non-object tool-call arguments instead of crashing (#1370)

A native function/tool call whose `arguments` field is valid JSON but not an
object — a bare array like ["ls -la"], or a string/number/bool/null — parsed
fine in function_call_to_tool_block and then every branch called args.get(...),
raising AttributeError ('list'/'str' object has no attribute 'get'). That
propagated out of the streamed agent loop (no surrounding try/except at the
call site in stream_agent_loop) and aborted the user's entire turn. Weaker and
local models routinely emit malformed args like this.

Coerce non-dict parsed arguments to {} (mirrors the existing empty-arguments
behavior), so the tool runs with empty args instead of killing the stream.

Adds tests/test_function_call_non_object_args.py covering array/string/number/
bool/null arguments — they fail before this change and pass after.

src/tool_schemas.py

@@ -1074,6 +1074,14 @@ def function_call_to_tool_block(name: str, arguments: str) -> Optional[ToolBlock
         logger.error(f"Failed to parse function call arguments for {name}: {arguments}")
         return None
 
+    # Some models emit valid JSON that isn't an object (e.g. a bare array
+    # ["ls -la"], string, or number) as the function arguments. Every branch
+    # below assumes a dict and calls args.get(...), so a non-dict would raise
+    # AttributeError and abort the whole agent stream. Coerce to {} instead.
+    if not isinstance(args, dict):
+        logger.warning(f"Non-object function call arguments for {name}: {args!r}; treating as empty")
+        args = {}
+
     tool_type = _TOOL_NAME_MAP.get(name, name)
 
     # Allow MCP tools through (namespaced as mcp__serverid__toolname)