Honor HEROKU_HTTP_PROXY_* settings when downloading slug from S3

[jgwhite]

Why the change?

Currently this command honors HEROKU_HTTP_PROXY_ env vars when it talks to the Heroku API but not when it downloads the slug from S3. This PR adds proxy support to the download step.

How do I verify this?

1. Check out the branch
2. Link it to your heroku CLI:

   heroku plugins:link <path-to-repo>

3. Boot up a proxy of some kind (I used mitmproxy)
4. Download a slug using the proxy:

   HEROKU_HTTP_PROXY_HOST=<your-proxy-host> \
   HEROKU_HTTP_PROXY_PORT=<your-proxy-port> \
   heroku slugs:download -a <your-app> <slug-uuid>

5. Verify that the slug downloads correctly and requests go via the proxy
6. Download a slug without the proxy:

   heroku slugs:download -a <your-app> <slug-uuid>

7. Verify that the slug downloads correctly

How do I do all that with mitmproxy?

1. Install mitmproxy:

   brew install mitmproxy

2. Boot it up:

   mitmweb --set stream_large_bodies=4k

3. Open System preferences and click your way to Network > Advanced > Proxies > Web Proxy
4. Enable “Web proxy (HTTP)” and set it to localhost:8080.
5. Visit http://mitm.it
6. Download the certificate but don’t worry about the rest of the instructions (these don’t apply for Node)
7. Disable “Web proxy (HTTP)”
8. Download a slug using the proxy:

   NODE_EXTRA_CA_CERTS=<path-to-the-cert> \
   HEROKU_HTTP_PROXY_HOST=localhost \
   HEROKU_HTTP_PROXY_PORT=8080 \
   heroku slugs:download -a <your-app> <slug-uuid>

9. Verify that the slug downloads correctly and requests go via the proxy
10. Download a slug without the proxy:

    heroku slugs:download -a <your-app> <slug-uuid>

11. Verify that the slug downloaded correctly

Why not use global-agent?

global-agent expects a single env var (with host and port) whereas the existing heroku-client expects two separate env vars. It seemed overall simpler to rig things up to match the existing heroku-client behavior.