After merging in layered tables, the CURSOR_API_CALL macro now takes a dhandle argument rather than a btree argument. We don't always have a dhandle available, so there's now a bunch of code that does stuff like CURSOR_API_CALL(..., btree == NULL ? NULL : foo). We should clean this up, possibly with another helper macro, or wrappers for CURSOR_API_CALL (e.g. CURSOR_API_CALL_SESSION_DH might use session->dhandle).