Ensure design docs are uploaded individually when replicating with bulk_get

[nickva]

Previously, when replication jobs used _bulk_get they didn't upload design docs individually like they do when not using _bulk_get.

Here were are preserving an already existing behavior which we had in the replicator without _bulk_get usage for more than 2 years. It was introduced here in #2426. Related to these issues #2415 and #2413.

Add tests to cover both attachments and ddoc cases. meck:num_calls/3 is helpful as it allows to nicely assert which API function was called and how many times.

[rnewson]

why?

There's no reason I can see to artificially limiting how many design documents are posted in one _bulk_get. Other clients are under no obligation to do the same, so any assumption you might make elsewhere, that ddocs are updated individually is built on a shaky foundation.

Please explain why this is an improvement, or bug fix, for couchdb on its own terms.

[jaydoane]

This seems like an elegant way to prevent using replication to exceed ddoc limits.

[nickva]

@rnewson good question, I should have explained that a bit better in the commit.

That is to preserve an already existing behavior which we had in the replicator without _bulk_get usage for more than 2 years. It was introduced here: #2426 related to these issues #2415 and #2413

EDIT: updated commit and PR comment with the references as well

[nickva]

Updated with the minor optimization and added references in the commit message and PR comment to the previous PR which introduced this behavior and the related issue.

[rnewson]

@nickva thanks for the links. It still seems to be a (perhaps necessary) hack around a problem elsewhere. We should do The Right Thing for any _bulk_doc write that writes multiple design documents. It is not right imo to just teach our replicator to avoid triggering those bugs, no other client will be similarly protected.

[nickva]

@rnewson with the replicator we are trying to be compatible even with old, or possibly broken clients. But most of all, in this PR, we are trying to bring back the previous behavior before the recent _bulk_get optimization. Currently, it's odd that the there is a difference based on whether [replicator] use_bulk_get is toggled or not.

So agree, we should do the right thing, but it should be in a separate PR. If we don't want this behavior we'd rip it out regardless of _bulk_get feature usage.