API Tweaks: Database.view(), Database.save_many()

[Jason Gerard DeRose]

Some misc API changes I grouped up here:

* Database.view() now includes reduce=False by default, which is really how the CouchDB API should work in the first place. For code that only uses a non-reduced view, it sucks for this code to have to know whether there is a reduce function. James and I discussed this a while back, so I'm trying it first in Microfiber, and then will add it to couch.js if we like it.

* Renamed new "non-atomic" Database.bulksave() to Database.save_many() because combined with Database.get_many(), it makes for a clearer, more consistent API.

* Renamed original "all-or-nothing" Database.bulksave2() back to Database.bulksave(). This way there is no API breakage, just in case anyone out there has been using this and wants the "all-or-nothing" behavior, understanding the implications. But I think we should deprecate this method soon. My idea with the convenience methods isn't to capture every scenario, but instead to just capture a few important patterns that might be tricky to get right using post(), get(), etc.

* Adds experimental id_slice_iter() function to give you all the ids from the rows is a view result, but chunked into groups of 25 (override with size=10 or whatever). This is for certain operations in Dmedia where we need to update a large number of docs, like the new Core.purge_store() method. So the idea is as you go through the view results, you use Database.get_many() to get 25 docs, update them all, and then save them back with Database.save_many(). This provides a huge performance improvement over get()/save() for each doc... 10x easily.