Instruction-level granularity in callgrind output

When generating callgrind format output, produce cost lines at
instruction granularity. This allows visualizers supporting the
callgrind format to display instruction-level profiling information.

We also need to provide the object file (ob=) in order for tools to find
the object file to disassemble when displaying assembly.

We opportunistically group cost lines corressponding to the same
function together, reducing the number of superfluous description lines.
Subposition compression (relative position numbering) is also used to
reduce the output size.

Ignore system name when merging nodes

This will allow merging nodes for multiple instantiations of
C++ templates. A graph option is introduced to preserve older
behavior.

Speed up proto postprocessing phase in proto decoder

This is a simple if/else logic so I agree an explicit else clause
is appropriate.

The proto decoder associates creates pointers to reference each
location/function/mapping based on their id. These ids can be
arbitrary uint64s, but often they are generated in sequence from
1 to N.

The overhead of keeping these indices in a hash is about a 20% of
the cost of decoding a profile. Speed it up by using an array to
track values from 1 to N, and a hash for values outside that range.

Disambiguate names for kcachegrind under the call_tree option

Speed up profile encode/decode

When using the call_tree option and generating a graph for kcachegrind,
it will merge back nodes that are distinct on the tree, producing some
confusing results. Add a suffix so that these entries are kept separate.

This addresses the problem described in
http://yosefk.com/blog/how-profilers-lie-the-cases-of-gprof-and-kcachegrind.html ,
particularly the summary "Choosing a profiler is hard" section.

Preallocate arrays to avoid repeated growth during profile encode/decode

Add symbolization support for Mach-O

Fix tree trimming

This will enable symbolization support for Go on Mac OS
It re-enables symbolization using debug/pprof/symbol on
Go profiles in the legacy format, and implements basic
mach-O support on the binutils package.

This is to keep the new TrimTree functionality from breaking any code
currently using the public interface. We do this by separating NodeSet
from nodePtrSet and creating different functions for each.

If we are deleting a node N, if the edge between N and its parent and
the edge between N and its child are both inline, then the resulting
residual edge from N's parent and the child will also be inline.

In a graph, NodeInfo maps one to one to the nodes, so it suffices to
just find the top NodeInfo s and only keep those nodes in the graph. In
a tree however, a single NodeInfo may map to many nodes. As of this
commit, a call to 'web 10' in pprof on a tree will return all the nodes
corresponding with the top 10 NodeInfo s.

Makes the report label accurate

Removing this file does not break any tests, though removing other cppbench.*
files do.

It is no longer representative of how omissions are made. pprof uses
more than just the cumulative value to make omissions.

Before this change, the node count used in the label is the proposed amount
provided by the user. If some nodes were trimmed and the graph ended up with
less nodes than the user asked for, the report label will now reflect this.

Allow weblist to work even if assembly is not available

Weblist provides source and assembly combined in a web document.
It needs access to the binary to print the assembly, but currently
refuses to generate the source if the binary can't be find.

Fall back to just generating the source if the binary isn't found.

Split source filepath cleanup from file opening

Reports want to use trimmed paths, not the location in the current
filesystem which may be very verbose.

Add mechanism for extensions to set variable defaults

This enables pprof extensions to update the default for a variable
before the command line is processed.

Add source_path option to point pprof to source files

Currently pprof will look for source files only on the current directory
and its parents. This makes it hard to examine sources on jobs where
there are multiple source trees (eg from different libraries).

Add a variable to provide a search path for source files. It will default
to the cwd, so there will be no change in behavior by default.

Generate kcachegrind reports under line granularity

Callgrind reports are generated with line granularity since kcachegrind
can take advantage of the information. However, kcachegrind reports were
not being generated with that granularity, creating an unnecessary difference.

Make Node sorting check all Info fields

Graphs are built using Go maps which do not provide a determinist
traversal. We sort the nodes to remove the non-determinism, but
the sort fails to provide a fully deterministic ordering in cases
where the attributes being sorted have equal values. Detect those
cases and fallback to a full comparison of all fields.