We recently ran into an issue on a small percentage of nomad-clients
where the nomad-client was running successfully, but due to a race
condition, could not correctly bind to the docker socket. This caused
all of our nomad jobs to be allocated to a single nomad-client instead
of being spread evenly across our clients. The only way to discover this
was to run `nomad node-status <node>` and count each job allocation per
node.

This can lead to a fairly long debugging process if there are several
nomad-clients. Including the number of allocations for each node in the
`node-status` command would save a large amount of debug time.

```
jake@biscuits [12:08:41] [~]
-> % nomad node-status
ID        Datacenter  Name      Class   Drain  Status  Allocations
2b0aabc5  dc1         biscuits  <none>  false  ready   0
```

```
jake@biscuits [12:08:55] [~]
-> % nomad node-status
ID        Datacenter  Name      Class   Drain  Status  Allocations
2b0aabc5  dc1         biscuits  <none>  false  ready   1
```

- Introduces a travis target in Makefile to update docker and
  install rkt

Lifted from https://github.com/fsouza/go-dockerclient

BSD Make uses `$.CURDIR` instead of `$CURDIR`. BSD and GNU share `$PWD`
though, so it works as a drop in replacement.

Also update scripts in `scripts/` to call `/usr/bin/env` in the shebang,
as BSD places `bash` at `/usr/local/bin/bash` instead of `/bin/bash`

These two tools are used in the "cov" Makefile target, but installing
these were not automated. This PR removes the ambiguity where a
developer tries to do a `make cov` and finds that the tools are not
installed.