Michael, would you please review and pull if you agree with these changes?

I've added a few specific comments in the diff; additionally, the following may be interesting:

• I've split __wt_session_get_btree into two functions, one of which checks the configuration strings for "checkpoint" entries, and one of which does the actual work; it might be worth a brief review of the places I call them both to be sure I didn't make a mistake, also, this set of changes collided with your recent changes to how table files are opened, so there's some question.

• FYI, there was a bug in the previous code, we didn't refuse to allow named checkpoints in the presence of an open backup cursor.

• Please skim txn/txn_ckpt.c, that's where the semantic changes are, specifically check if the loop to lock the checkpoints being deleted looks correct to you.