backtrace semantics

[David Miller]

First, thanks for adding all the new backtrace() test cases.

There seems to be an inconsistency about how to handle the top-most
frame in the backtrace.

x86 for example, includes the caller of backtrace(), and this matches
the expectation of the new testcases.

Whereas s390 and sparc, for example, elide the backtrace() caller's
frame from the backtrace, and starts reporting at the frame
immediately afterwards.

Which one is correct?