Clarify that a switch expression may have no type

[jskeet]

Fixes #1651.

I don't think we need to talk about "if a switch expression is not subject to..." - we just say that the expression has no type, but can be converted to some types, and let the rest of the language rules handle it.

We may want to add examples, e.g.:

• A switch expression with a type being used to initialize a var variable
• A switch expression without a type being used to initialize an explicitly-typed variable
• A switch expression without a type failing to be used to initialize a var variable

I don't know whether those examples are really needed though.

[Nigel-Ecma]

I think this approach fits into the existing approach to ”no type”.

I would probably myself drop lines 3882-3 as they just repeat what is said in the xref and instead add the xref to line 3885 – more succinct, but that is subjective!

[jskeet]

I would probably myself drop lines 3882-3 as they just repeat what is said in the xref and instead add the xref to line 3885 – more succinct, but that is subjective!

Makes sense - have done that in another commit (as well as reverting an accidental incomplete change for a different issue).

[BillWagner]

This LGTM as well.