[Suggestion]: Add --strictReadonlyChecks option to catch runtime errors modifying frozen objects at compile time

[paulkoerbitz]

The Problem
It is currently not possible to consistently catch modifications of frozen objects, which lead to runtime errors in strict mode. See #11180 and #11480 for examples.

Proposed Solution
Add a strictReadonlyChecks compiler option, which does not allow to assign readonly fields to non-readonly (regular) fields. Also do not allow deleting items from readonly dictionaries (#11480), unless this can already be solved by regular compiler behavior.

[RyanCavanaugh]

Having this in any usable state would require a full set of variance and mutability annotations along with corresponding type operators -- a very large change. We're thinking about a future with some kind of variance / mutability annotations but won't be doing this as a one-off feature.

[DavidANeil]

I think there are 2 distinct things that I see in the issues referencing this one.

1. A need for a readonly member to not be assignable to a readwrite member.
2. A need for a readwrite member to not be assigned to a more inclusive type unless it is also marked as readonly.
   Would it be reasonable to address these issues seperately?