i#4083 func trace: Add func_view tool

[derekbruening]

Adds a new func_view tool for visualizing and analyzing function calls
recorded in a drcachesim trace. It does two things in its initial
form: it reports the summary count of calls and returns for every
function traced, and if a new option op_show_func_trace is set (on by
default), it prints a linear sequence of every call and return with
return addresses, argument values, and return values. Each is
indented to enable a clear picture of the call sequence.

Adds a new option op_funclist_file to specify the file's path, though
normally it will be auto-found.

Adds raw2trace_directory_t support for parsing the funclist file.

Adds example output for the new tool to the docs. Adds a brief
section to the docs on function tracing in general, which was
previously missing.

Adds a func_view test which runs fib(5) and ensures we get the nested
function trace correct.

Fixes #4083

[derekbruening]

Well I got the test to pass on my local Windows machine, but on Appveyor it is hitting a mysterious crash with no further information:

[00:24:03] 205: Running postcmd |C:/projects/dynamorio/build/build_debug-internal-64/clients/bin64/drcachesim.exe;-indir;C:/projects/dynamorio/build/build_debug-internal-64/suite/tests/tool.drcacheoff.func_view.common.fib.exe.07132.7209.dir;-simulator_type;func_view|
[00:24:03] 205: CMake Error at C:/projects/dynamorio/suite/tests/runmulti.cmake:106 (message):
[00:24:03] 205:   *** postcmd failed (Access violation): ***
[00:24:03] 205: 
[00:24:03] 205: Call Stack (most recent call first):
[00:24:03] 205:   C:/projects/dynamorio/suite/tests/runmulti.cmake:118 (process_cmdline)
[00:24:03] 205: 
[00:24:03] 205: 
[00:24:03] 205/239 Test #205: code_api|tool.drcacheoff.func_view ...........................***Failed   14.80 sec

A crash is disturbing. But If I can't reproduce I will probably have to disable the test on Windows and put this PR in and consider the crash separately.

[derekbruening]

Hmm, Mac is starting to have strange results. Twice now the log is truncated, but there's no timeout message, and the time is not some even number.

[derekbruening]

A64 is #3742 threadexit2