plugins/driver: update driver interface to support streaming stats

client/tr: use streaming stats api

TODO:
 * how to handle errors and closed channel during stats streaming
 * prevent tight loop if Stats(ctx) returns an error

drivers: update drivers TaskStats RPC to handle streaming results

executor: better error handling in stats rpc

docker: better control and error handling of stats rpc

driver: allow stats to return a recoverable error

tests: remove tests for unsupported features

With switching to driver plugins, driver validation is quite tricky and
we need to do some design thinking before supporting it against.

added validation on client metadata keys

Improve Kill handling on task runner

Co-Authored-By: dadgar <alex@hashicorp.com>

drain: fix node drain monitoring

tests: run docker tests in macOS out of box

Extract LXC from nomad

drivers: avoid referencing client/structs package

Add another task events e2e test and fix test helper

Fix a panic on node.Deregister fail

Fix Travis MacOS job 

This PR improves how killing a task is handled. Before the kill function
directly orchestrated the killing and was only valid while the task was
running. The new behavior is to mark the desired state and wait for the
task runner to converge to that state.

node.NodeResources is nil when operating against pre-0.9.

vendor: Update go-plugin

Before it was *exactly* N nodes which limited test portability between
clusters.

validation in the future

config dir clarification

Use `/tmp` as temporary directory for docker driver tests, so tests can
run out of the box without any intervention.

macOS sets tempdir as `/var`, which Docker does not whitelist as a path
that can be bind-mounted.

drivers/exec: restrict devices exposed to tasks

Allow scheduling mock job when running on MacOS (or Windows) hosts.

The whole approach to monitoring drains has ordering issues and lacks
state to output useful error messages.

AFAICT to get the tests passing reliably I needed to change the behavior
of monitoring.

Parts of these tests are skipped in CI, and they should be rewritten as
e2e tests.

Emit Killing task events and add e2e tests

Error expects an odd number of arguments, and panics otherwise.

This fixes a bug related to shutting down of GRPC plugin interfaces
(more info: https://github.com/hashicorp/go-plugin/pull/88)

This does not yet fix all test cases for subprocess leaking, but is a
useful independant change.

test: re-eanble periodic fingerprint test