Readonly members more test coverage

[RikkiGibson]

This PR contains test coverage additions mainly from the test plan review we had yesterday.

• Check parser behavior for bad readonly property int P readonly => 42;
• Improves warning message for implicit value copies
• Require both property accessors be declared to use readonly keyword on an accessor (this matches the behavior with accessibility modifiers)

This is the last work I expect to do before finally merging master into this branch, and then integrating this branch into master.

[jcouv]

This is the last work I expect to do before finally merging master into this branch

Did you do an IDE test pass to identify what works and what doesn't? It's generally ok to file issues, except for few blockers: mostly crashes, annoying completion (ie. you can't type what you want and have to fight against the IDE), or any really annoying behavior.

[RikkiGibson]

I've played with the feature in the IDE, authoring and referencing readonly members similar to what's written in the speclet, and found it worked pretty smoothly.

What's missing is generally what I've listed under the IDE section in #32911. Maybe that section should be migrated to a new issue?

[jcouv]

Sounds good.

Maybe that section should be migrated to a new issue?

I would file individual issues. That isn't blocking the merge.

[jcouv]

LGTM (iteration 5) with one test suggestion

[RikkiGibson]

I'm adding in a minor rewrite of an error message as discussed with @cston and @agocke.