Fix stream cleanup timeouts
The first part of this is adding the `after 0` clause. The issue here is
that ibrowse sends the `ibrowse_async_response_end` message without
waiting for a call to `ibrowse:stream_next/1`. This means that the
continuous changes feed may or may not get this message in
`couch_replicator_httpc:accumulate_messages/3`. If it does then we would
end up on an infinite timeout waiting for it. This was a typo in the
original patch in that I meant to include it but forgot.
The second timeout is so that we don't end up halted waiting for a
changes request to finish. If it takes longer than 30s we just crash the
replication and let the manager restart things.
BugzId: 47306
This is a cherry-pick of:
https://github.com/cloudant/couch_replicator/commit/2caf39040e4e50c38a7758d4d09e7a8b22ea92d4
1 file changed
