Make bulkUpdate non atomic function

[ssh24]

Description

Make bulkUpdate a non atomic function. Allow connectors to perform bulkUpdates now.

connect to loopbackio/loopback-connector-cloudant#35

[bajtos]

@ssh24 Could you please explain what problem is this pull request trying to solve?

bulkUpdate API is an integral part of change replication/offline sync implementation. I am concerned that by moving this implementation down to connectors, we make it easy for 3rd-party/community connectors to break change replication by implementing an incomplete version of bulkUpdate. To address that, I would like to see a robust test suite covering bulkUpdate in juggler's shared test suite run by all connectors.

I also think this is a breaking (semver-major) change - existing applications will stop working after they update to a newer loopback version (with bulkUpdate removed) in case their connector does not provide its own bulkUpdate implementation. A possible solution for preserving backwards-compatibility is to keep bulkUpdate implementation inside loopback, mark it as deprecated, and let connectors to override this loopback-provided version with their own implementation, similarly how they are already overriding other DAO methods.

[bajtos]

☝️

Allow connectors to perform bulkUpdates now.

This should be already possible, using the same mechanism which overrides dummy "create" with "create" provided by DataAccessObject.

[ssh24]

@bajtos This PR and the subset PR's are not complete. Thank you for joining the conversation. I would like @raymondfeng and @kjdelisle to join as well.

PR in juggler: loopbackio/loopback-datasource-juggler#1399
PR in cloudant: loopbackio/loopback-connector-cloudant#150

The goal here was to allow cloudant connector perform bulkUpdate. This is definitely a breaking change in loopback. The tests are failing as well on bulkUpdate. So I want opinion on this and what would be the best way to do it.

[ssh24]

I realized we can override this function in DAO. Closing this PR.