C#: Support for record structs

[michaelnebel]

In this PR we make support for record structs and change the original implementation for record.
Characteristics of the old implementation

• When records were introduced in C# they were implicitly record types.
• The QL code used the presence of a method to decide, whether a type should be considered a record type. The method doesn't exist for record struct's, which means we can't reuse this idea to identify, whether a type is a record.

It appears that there is nothing in IL level that identifies, whether something is a record type (it seems native to C# only).
In C# 10, it is now also possible to declare a record struct and record class. record class is clarification of record (that is record is just a shorthand for record class). That is record implicitly means record class and as such, we can consider the record keyword to be a modifier that applies to class and struct.

[michaelnebel]

@tamasvajk @hvitved This is the initial suggestion for record struct support. I am not sure,

• if it is acceptable that we remove some of the functionality from cil
• if we should delete the RecordClone and friends classes.

[tamasvajk]

• if we should delete the RecordClone and friends classes.

I don't think we should delete RecordClone altogether. That's used as the target of a ... with { ... } expression and might be useful for query writers. BTW, we might need to change the WithExpr in QL to support the C# 10 changes. I added a separate point to the internal C# 10 issue.

[michaelnebel]

• if we should delete the RecordClone and friends classes.

I don't think we should delete RecordClone altogether. That's used as the target of a ... with { ... } expression and might be useful for query writers. BTW, we might need to change the WithExpr in QL to support the C# 10 changes. I added a separate point to the internal C# 10 issue.

Great - thank you! 😄

[hvitved]

Approach looks good to me.