Always keep a field for private[this] var class parameters

[lrytz]

Reads of var class parameters in the constructor are not rewritten to
the constructor parameter, they keep referring to the field to ensure
potential field writes are reflected (this was fixed in #7688).

This means that the corresponding field cannot be elided even if it's
never written.

The alternative solution would be to detect private[this] var fields
that are only read in the constructor. However this is difficult to
implement, see #9143.

The second commit goes a step further by never eliding the field
for a private[this] var class parameter. Up for discussion.

Fixes scala/bug#12002 (the remaining part).

This needs to be backported to 2.12.x.

[lrytz]

I'm personally in favor of the simpler solution, always keep the field. Even dotty agrees

➜  sandbox git:(t12002b) dotc ../test/files/run/t12002.scala && dotr Test
good boy!
List()
List()
List(private int C.x)
List(private int D.x)
List(private int E.x)
List(private int F.x)
List(private int G.x)
List(private final int H.x)
List(private int I.x)

[som-snytt]

On one of the PRs, I mention that doti doesn't do it, but it wasn't clear what "inferred private[this]" would entail in doti.

[som-snytt]

I agree there is not much benefit in this secret handshake. If user doesn't want a field, they should change their code.

The possible use case was if class var param was mutated only during constructor, but otherwise unused.

The test fail is that neg/t1503 with Nil.type.

[lrytz]

The possible use case was if class var param was mutated only during constructor, but otherwise unused.

👍 but you can't do that with method parameters either, so no big loss and in a way it makes things more regular