Add Kubernetes metadata

[axw]

Add Kubernetes metadata. We sniff pod metadata from /proc/self/cgroup, and then add or override any values specified through the environment variables KUBERNETES_NAMESPACE, KUBERNETES_NODE_NAME, KUBERNETES_POD_NAME, and KUBERNETES_POD_UID. These can be populated using the Kubernetes Downward API.

Through /proc/self/cgroup we are able to sniff the container ID and pod UID. We also assume that the hostname is the pod name, which is the default. If the hostname has been explicitly set, then the user can use the Downward API as mentioned above to set the correct pod name.

Closes #329

[codecov-io]

Codecov Report

Merging #342 into master will increase coverage by 2.4%.
The diff coverage is 40.38%.

@@            Coverage Diff            @@
##           master     #342     +/-   ##
=========================================
+ Coverage   81.14%   83.54%   +2.4%     
=========================================
  Files          98       98             
  Lines        5891     6009    +118     
=========================================
+ Hits         4780     5020    +240     
- Misses        840      923     +83     
+ Partials      271       66    -205

Impacted Files	Coverage Δ
model/marshal_fastjson.go	75.12% <0%> (+2.15%)	⬆️
internal/apmhostutil/container.go	100% <100%> (ø)	⬆️
utils.go	80.68% <44.82%> (-16.1%)	⬇️
internal/apmhostutil/container_linux.go	85.18% <85.18%> (ø)
modelwriter.go	94.16% <0%> (+0.83%)	⬆️
internal/radix/radix.go	39.92% <0%> (+0.92%)	⬆️
internal/sqlscanner/scanner.go	96.82% <0%> (+1.58%)	⬆️
metrics.go	91.66% <0%> (+1.66%)	⬆️
... and 34 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0ac0db8...3c7e66d. Read the comment docs.

[axw]

While testing, I realised that the container.id wasn't being populated. That's because the cgroup paths are different under k8s. They look like this:

1:name=systemd:/kubepods/besteffort/pode9b90526-f47d-11e8-b2a5-080027b9f4fb/d87d73916efef57277e02f38eea58495a5e3687a494e34ffa8e14fc2b8a52dc3

The last segment is the Docker container ID. Obviously only holds if Docker is the container runtime being used. The second-to-last segment is the pod UID, prefixed with "pod". I also realised we can get the pod name just from the hostname.

Need to run some more tests, with the following variables:

• kubelet cgroup driver: cgroupfs vs. systemd
• alternative container runtime?