Delivery: Create Sourcegraph OS image

[abeatrix]

Suggested by CE @christinelovett: create a single Sourcegraph OS image that can be used across different cloud provider as well as on perm customers.

More content from @jdpleiness on Slack:

I have a few ideas that may or may not be useful. I think using a single OS for the images would cut down maintenance. Right now the Amazon image is using Amazon linux, while GCP and Azure are using Ubuntu. Perhaps switching all images over to Ubuntu after initial rollout could be something to look into. Another idea I have been thinking about is creating a “sourcegraph daemon” of sorts that would run on the images. While initially this might not seem super useful (mainly replacing the reboot.sh script functionality) I think long term having a daemon running would give us a platform for other automation we might want to include such as health checks, simple fixes that can be automated, upgrades perhaps, etc.. I could also see this connecting to a CLI interface that could help with troubleshooting/setup tasks if needed.

[slimsag]

One concern / thing to look into here:

• Do the various marketplaces have restrictions on base OS? i.e. does Amazon marketplace require/highly suggest we use Amazon Linux? I'll ask Tom Pinckney
• Will using the non-default image raise any security questions for any enterprise customers? Or is Ubuntu standard enough that it'll be fine? Could probably ask Diego for his opinion here.

[beyang]

I think many enterprise customers may have a requirement to use Red Hat Linux (aka Centos, Amazon Linux 2) over Debian/Ubuntu in prod. I imagine among cloud deployment, some flavor of RHEL is a more common default than Debian. Is there a customer-driven reason we chose Ubuntu for GCP/Azure? If not, I almost feel that some flavor of RHEL would be better for those (and perhaps easier to maintain, since there would be less divergence, if any from the AL2 image).