Codegen ignores Secret flag on properties with prompt types

[t0yv0]

Hello!

• Vote on this issue by adding a 👍 reaction
• If you want to implement this feature, comment to let us know (we'll work with you on design, scheduling, etc.)

Issue details

Currently in providers such as AWS we have properties that are specified as secret, but are projected by codegen into promptly typed (not Output<T>-wrapped) accessors. As a result there is no guarantee that consuming code understands they are secret. We should fix this, including a breaking change to project these properties as Output<T>-wrapped properties so we can set the secret flag on them.

Since #414 this is a codegen warning but we can tighten to an error and fix for practical cases.

Affected area/feature

[t0yv0]

NOTE to self: compare to what happens in other languages to decide what to do here.

[t0yv0]

The underlying issue promoted to pulumi/pulumi#9744 as it affects more than just Java. For this issue I suggest we close with the attached PR that tightens the warning for now.