gRPC and HTTP default views that break down latency by status

[semistrict]

Currently our gRPC and HTTP views only break down latency by method. I would like to propose adding a new view to each of {gRPC,HTTP} {client,server} default views that also breaks down by status (in addition to method). It's important for SLIs that we know the latency of successful vs. unsuccessful requests.

[bogdandrutu]

Sounds very reasonable. Please create a PR.