Reusing a compiler instance with -Yinline causes an inconsistent state

[scabug]

If the REPL or fsc is used with -Yinline then we end up in a state where two symbols represent the same thing and the second version isn't initialized where it's expected. This is the root cause for the linked bugs which have more discussion on the subject.

Undoing the fix for #SI-5789 creates a simple reproducer. The test associated with that fix will cause an assertion in Mixin.scala.

[scabug]

Imported From: https://issues.scala-lang.org/browse/SI-6782?orig=1
Reporter: @JamesIry
Affected Versions: 2.10.0
See #5776, #5789, #6177

[scabug]

@retronym said:
WIP: https://github.com/retronym/scala/tree/ticket/8871

[scabug]

@retronym said:
I've saved my attempt to fix this in https://github.com/retronym/scala/tree/ticket/6782. Unfortunately it causes a regression in inlining of trait methods under separate compilation.

% qscala -nc -optimize -Yinline-warnings
Welcome to Scala version 2.11.3-20141010-000123-a654060c93 (Java HotSpot(TM) 64-Bit Server VM, Java 1.6.0_65).
Type in expressions to have them evaluated.
Type :help for more information.

scala> "".stripMargin
<console>:8: warning: Could not inline required method augmentString because bytecode unavailable.
              "".stripMargin
              ^
<console>:8: warning: At the end of the day, could not inline @inline-marked method augmentString
              "".stripMargin
              ^

[som-snytt]

The simple reproducer was lost to new traits in 2.12.