Transient memcache messages being reported by the CLI

[lblackstone]

After #348, the kubernetes-provider is printing info messages from the cached discovery client for some transient failures. For example:

pulumi:pulumi:Stack (eks-test-dev):
    E0204 16:05:50.257074   17939 memcache.go:126] couldn't get current server API group list; will keep using cached value. (Get https://96059DADE46C784B628BDBDDC8F0CD8A.sk1.us-west-2.eks.amazonaws.com/api?timeout=32s: dial tcp 54.71.80.10:443: i/o timeout)

These messages are not actionable and the underlying error should be automatically resolved by the retry logic, so we need to figure out a way to disable them. It appears that the underlying library is bypassing the CLI's logger, and writing directly to stderr.

[lblackstone]

Here's an example of a successful update that printed this warning message