Don't fail on client for Batch API for large requests

[abhijitpai]

Description

We had an optimization to not make Batch requests to the server if we think they are too large - removing this since making server requests allows for the natural error handling flow (similar to regular point operations where we don't check the max size of input on the client) among other benefits.

Type of change

• Bug fix (non-breaking change which fixes an issue)

[ealsur]

Just one question. In the case of Bulk, if the operation is too big, would then the Batch partially fail with a failure for that operation? If so, that means that operation can never go through Bulk/Batch due to its size, correct? Which is the expected server status response?

[abhijitpai]

@ealsur
Since we will no longer restrict the resource size within the batch operation on the client, if we have a resource with size beyond resource size limit but the batch request itself is within the max batch size limit, then we will go to the server and the batch request can fail partially, with the particular operation result's status code being HTTP 413 - I am adding an emulator test in bulk to demonstrate this behavior.

In the earlier iteration, I had tried to keep rest of the bulk flow as it was before this change since bulk needs to know about batch request size limit on the client in order to be able to dispatch correctly unlike atomic batch so we could take any needed fixes in a separate PR; I will instead fix them in this PR in the next iteration so there is less confusion and we will never block on the client for large operations within bulk as well - the client size limits we have still apply in bulk except that we will ensure the server batch request has at least 1 operation irrespective of size configuration.

[ealsur]

Thanks!