Docs: assume usage of helm3 over deprecated helm2

[GeorgianaElena]

To do list (based on @consideRatio's summary here ✨ ):

• no need to --purge
• new --timeout spec
• --cleaup-on-fail always when upgrading
• no tiller
• make it clear that the docs assume helm3
• instructions on how to migrate from helm2 to helm3 for an existing cluster
• remove section about helm2 setup (or keep it for compatibility)

[mriedem]

FWIW we recently upgraded our testing environment hub from helm2 to helm3 and one of the things we needed to do was set c.JupyterHub.hub_connect_ip = 'hub' so that notebooks hit the hub API URL by hostname rather than service IP (which changed with our upgrade to helm3).

[manics]

I've made some suggestions after a first pass over the docs.

[consideRatio]

@GeorgianaElena thank you for working on this and thanks @manics for the review! ❤️

For instructors on how to migrate from Helm2 -> Helm3, I think we should save ourselves some work by referencing the official docs straight up even though I suggested otherwise before. They are complicated to follow, but I don't think we should spend time trying to simplify them.

[GeorgianaElena]

Thanks for reviewing this @manics and @consideRatio ❤️

[consideRatio]

Wieee! Thanks for your work on this @GeorgianaElena!

I'll go ahead and merge even though there is an error in CI because i know it is transient and fixed in current master.