Add additional cloud provider driver "helper" functions

[atheriel]

The odbc::databricks() and odbc::snowflake() functions have been pretty successful, because we've been able to stuff a lot of unpleasant auth and configuration details in there that users would previously have to copy around manually.

There are several other obvious candidates for "cloud provider helper functions", too:

• redshift() and rds() helpers that handle AWS IAM credentials and maybe integrate with the paws package; and
• A sqlserver() helper that picks up on Azure credentials.

I think we should try to build these out over time.

[hadley]

I think we should do this for all databases, since it's often not easy to figure out the needed parameter names.

[atheriel]

Something we could do for non-cloud provider databases in those helper functions would be system keyring support, and better hygiene around passwords in general.

I think right now code snippets suggest doing things like pwd = rstudioapi::askForPassword(), but we don't really formalise that advice like we can with the helper functions. E.g. I can imagine code omitting a password like the following:

dbConnect(odbc::postgres(), server = "pgsql.internal", uid = "aaron")

would call the equivalent of

keyring::key_get("psql.internal", username = "aaron")

and then finally askpass::askpass() in interactive contexts.

[hadley]

Yeah, we just need to figure out which package should be responsible for this infrastructure and design some conventions around it. I suspect that this should be keyring's responsibility .