Handle frame pointer omission about google-breakpad HOT 17 CLOSED

I'm starting with a bite-sized piece.  This structure is used to store frame 
data
information in the processor.

This is another bite-sized piece.  This is the stack machine that's used to 
evaluate
program strings, which is how we'll be able to use stack frame data to 
determine how
to get from one frame to the next in the absence of frame pointers.

ContainedRangeMap v2

ContainedRangeMap v2 checked in, revision [29].

PostfixEvaluator checked in, revision [32].

SourceLineResolver and PDBSourceLineWriter changes.  This augments the symbol 
file
format to carry the data we need.

I've actually got the rest of the FPO stackwalking stuff working, tied in to
PostfixEvaluator in StackwalkerX86.  I'm not posting that yet because I'm still
trying to figure out the best way to store context, and it might require 
subclassing
StackFrame or doing other invasive work.

New test data for the SourceLineResolver and PDBSourceLineWriter changes.

SourceLineResolver and PDBSourceLineWriter changes, updated per review comments.

The test data in dump_syms_regtest.out was from an older version of
PDBSourceLineWriter and will cause tests to fail.  I'm not going to attach a new
testdata patch, but when I check this in, I'll check in the correct version of 
the
test data.

SourceLineResolver and PDBSourceLineWriter changes checked in, revision [38].

Should I file a separate bug for tracking stackwalks across code that uses 
alloca or
C99 local arrays? I don't know that it's broken, but we should at least have
testcases for it.

Yes, feel free to file bugs for any areas you think we could improve test 
coverage.

Status update: I finished the initial implementation of the FPO stackwalker last
week, but found it to be buggy in testing.  I've now come up with a full set of
heuristics that should let us walk any optimized-code Windows x86 stack as well 
as,
if not better than, the native debugger.  I figured out what was missing by 
tooling
around in WinDbg (instead of just the VC debugger), which helped show me that I 
had
been misled by two bugs: one in my own code, and one in Microsoft's pdb dumper 
tool
dia2dump.

FPO stackwalker and associated changes.  There are a number of interface 
changes in
this patch.

Updated test data to accompany airbag.21.fpo.1.patch.

Addresses review comments.

The whole thing's fixed and checked in, revision [42].