update protobuf lib to the latest release 3.6

[mikejiang]

The current bundled pb is 2.6, which is 4 year old. We will have to test the backward compatibility (for both code and data archive) before making such upgrade.

[mikejiang]

In file included from /Users/Mike/Library/R/3.6/library/RProtoBufLib/include/GatingSet.pb.h:22:
In file included from /Users/Mike/Library/R/3.6/library/RProtoBufLib/include/google/protobuf/generated_message_util.h:44:
In file included from /Users/Mike/Library/R/3.6/library/RProtoBufLib/include/google/protobuf/stubs/once.h:81:
In file included from /Users/Mike/Library/R/3.6/library/RProtoBufLib/include/google/protobuf/stubs/atomicops.h:179:
/Users/Mike/Library/R/3.6/library/RProtoBufLib/include/google/protobuf/stubs/atomicops_internals_macosx.h:164:10: warning: 'OSAtomicAdd64Barrier' is deprecated: first deprecated in macOS 10.12 -In file in

The compiling warnings are flooding the flowWorkspace package building process, which could make it difficult to troubleshoot the real compiling errors.

[mikejiang]

pb is upgraded to 3.7 now in cytoset branch (with 6b96842).
Tthe proto message file is kept unchanged because switching to proto3 syntax will cause some breaking changes to the cytolib code due to the changing behavior of optional fields, i.e. the missing parent reference now returns a default 0 value instead of null, which will be ambiguous in our context since we interpret 0 as root node.

Preliminary tests show that it can successfully load the gs archived by pb2.6.
Right now the bundled pb includes everything and we can trim it down later if the speed of RProtobufLib package building ever becomes an issue.
Also the bundled windows dll can be updated later as well,but it is not urgent either since pb2.6 and pb3.7 play nicely with each other without any issues.

[mikejiang]

We can potentially further optimize the pb code footprint by only using protobuf lite, which will downsize the protocol buffers runtime significantly (libprotobuf.so 31M -- > libprotobuf-lite.so 4.8M)as well as reduce the interface overheads.

[mikejiang]

tests show protobuf-lite doesn't improve load_gs much. But since it requires more lightweight runtime, I will still use it.

[mikejiang]

We will hold off merging these to trunk until windows static lib for protobuf-lite3.7 is built and tested for cs branch.