Fixing Supervisor's PATH setting & two other treats

[fnichol]

This changeset has three parts:

[sup] Update how Supervisor sets PATH on start.

This change uses a more robust strategy for setting the PATH
environment variable before a supervised process is start via the
start subcommand.

Prior to this change, an install of the supervisor package was required
in order for the Supervisor to start any service, which included
running the Supervisor in the source tree for purposes of development.
This change makes it more explict what we are after in the supervisor's
PATH entry: a minmal BusyBox userland.

There is a series of fallback strategies used here in order to find a
usable BusyBox installation. The general strategy is the following:

• Are we (the Supervisor) running inside a package?
  • Yes: use the BusyBox release describes in our DEPS metafile
    & return its PATH entries
  • No
    • Can we find any installed BusyBox pacakge?
      • Yes: use the latest installed BusyBox release
        & return its PATH entries
      • No
        • Is the busybox binary present on $PATH?
          • Yes: return the parent directory which holds
            the busybox binary
          • No: out of ideas, so return an error after warning
            the user we're done

The end result is that an installation of the Supervisor package is no
longer required in a development environment--only an installation of
the core/busybox-static package. Additionally, the precise release of
the core/busybox-static dependency will be used when running the
Supervisor out of a pacakge (its natural home).

[formatting] Code formating updates due to rustfmt.

Without this change there will be mega code churn in the core::crypto module.

[core] Add env::var & env::var_os functions to check for empty vars.

The standard library will consider an empty-but-set environment
variable a legit result. After inspecting the codebase, it turns out
that is not the expectation in most of our code. Indeed, a set-but-empty
HAB_DEPOT_URL is not valid and should be ignored.

This change introduces 2 functions which shadow their standard library
cousins providing this additional check.

[thesentinels]

By analyzing the blame information on this pull request, we identified @metadave, @reset and @adamhjk to be potential reviewers

[fnichol]

It turns out that having an offline Depot is uncovering several of these issues which are a bit easier to fix. Don't ask me why…

[reset]

[fnichol]

@reset You have won the day for Gif That Makes Me Blush

[reset]

@fnichol I was very proud of that gif find