Configurable log level for AggregateFuture

[anjor]

Context
We use the whenAllSucceed method in Futures to combine futures when batching requests. This is done for the "fast failing" that AggregateFuture provides.

When more than one future throws an exception, one of the exceptions gets propagated, but the others get logged here at SEVERE level. For our specific use case, the exceptions don't matter as much because we have retry logic on the server side to handle them.

Problem
We have monitoring in place for logging at the SEVERE level. As a result of the above behaviour, our monitors keep firing despite the server handling the retry logic correctly.

Proposed Solution
Could we make the logging level configurable?

It will need to be passed from AggregateFuture to CombinedFuture up to Futures#whenAllSucceed. It could be made optional, and if it isn't set then it falls back to SEVERE to preserve backwards compatibility.

[cpovirk]

I have felt bad about this for years, but util.concurrent has remained not a priority for us for years, so I'm not sure when that will change.

You can follow along on #2134. I recently suggested a workaround that I haven't tested but that I am hopeful would work.