feat: Added matching engine integration

[ivanmkc]

Added matching engine integration.

Tasks

• Add Index and IndexEndpoint classes with create, get, list, update, delete methods
• Add IndexEndpoint.deploy_index() method that accepts Index object
• IndexEndpoint.undeploy_index
• IndexEndpoint.mutate_deployed_index
• IndexEndpoint.metadata.config parameters
• IndexEndpoint.undeploy_all
• Prediction/matching methods.

Testing

• Unit tests
• Integration test

Potential convenience methods

• MatchingEngineIndexEndpoint.delete(..., force: bool)

Note

MatchingEngineIndexEndpoint.match could not be unit tested due to https://stackoverflow.com/questions/47202355/pytest-cannot-mock-init-of-my-class.

It's been tested in the integration test.

Fixes b/201313324, #994

[ReneeZhuGG]

LGTM, left a few minor comments.

[morgandu]

metadata?

[ivanmkc]

Endpoint.undeploy_all doesn't have it so I followed that.

This loops through each. deployed_index and undeploys it.
Do you propose sending the same metadata for every deploy call?

[morgandu]

Thanks for the great work @ivanmkc , left some comments, mostly minor issue, except one pattern on the resource operation assuming resource is successfully created. Since the resource creation is optional_sync supported, any immediate operation on a creating resource may lead to error i.e. self.resource_name is not available yet. One option is add self.wait() in all downstream operations, this will be probably better if the resource creation takes time. Another option is remove the optional_sync from resource creation.