Same-type delegation breaks inlining: counted as recursive inlining

Originally reported at:
http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2012-September/008367.html

The benchmark is attached, and requires JMH to run.
This issue still reproduces perfectly well, for the same reason:

o.o.p.RecursiveBench.chain11: 4.4 +- 0.1 ns/op
o.o.p.RecursiveBench.chain12: 3.1 +- 0.1 ns/op
o.o.p.RecursiveBench.empty: 2.0 +- 0.1 ns/op

This issue also breaks the inlining in the case of JDK 8 Streams, in the scenarios like:
 LongStream.range(1, 10000)
   .filter(i -> i % 3 == 0)
   .filter(i -> i % 5 == 0)
   .sum();

...because we create the same wrapper classes for both filters.