[DNM] [Sema] Check whether wrappedValue initializer exists when generating …

[mininny]

Currently, CSGen assumes that init(wrappedValue:) exists in the definition of property wrapper parameter's type.
This is true when there's no other initializer in the definition as one will be synthesized. However, if there's init(projectedValue:) in the definition, the init(wrappedValue:) will not be synthesized.

Since the current code assumes that the init(wrappedValue:) exists, the compiler continues to apply the constraints. However, because it is missing, the compiler crashes at https://github.com/apple/swift/blob/94c2be87da7d17b94fa2c1f1418ea6d90f9af4b9/lib/Sema/CSApply.cpp#L6149C29-L6149C52 since the property wrapper itself is visible, but doesn't have an initializer that was actually applied.

---

Changes in this PR will fail the CSGen process if the init(wrappedValue:) is missing. However, without proper diagnostic, the error message is just shown as type of expression is ambiguous without a type annotation.

I think either one of these is possible:

• use init(projectedValue:) if init(wrappedValue:) doesn't exist but the initializer parameter type is identical.
• show a better diagnostic by recording a CSFix?
• leave it as type of expression is ambiguous without a type annotation since the compiler at least doesn't crash.

I want to show some better diagnostics like cannot use property wrapper argument because 'init(wrappedValue:)' doesn't exist.

• Am I allowed to produce the diagnostic by creating a CSFix like RemoveProjectedValueArgument?
• Or should I use some other pass of the compiler to produce the diagnostic, e.g. when typing checking the expression?

Resolves #68570

[taempark]

A new line, please.

[pcw109550]

A new line, please.